update ChangeLog

Change-Id: I39043d4986664312947a0668cb1b7bfbcf5a2477

AUTHORS

@@ -1,4 +1,5 @@
 Contributors:
+- Alan Browning (browning at google dot com)
 - Charles Munger (clm at google dot com)
 - Christian Duvivier (cduvivier at google dot com)
 - Djordje Pesut (djordje dot pesut at imgtec dot com)
@@ -9,6 +10,7 @@ Contributors:
 - Johann (johann dot koenig at duck dot com)
 - Jovan Zelincevic (jovan dot zelincevic at imgtec dot com)
 - Jyrki Alakuijala (jyrki at google dot com)
+- Konstantin Ivlev (tomskside at gmail dot com)
 - Lode Vandevenne (lode at google dot com)
 - Lou Quillio (louquillio at google dot com)
 - Mans Rullgard (mans at mansr dot com)
@@ -35,4 +37,6 @@ Contributors:
 - Urvang Joshi (urvang at google dot com)
 - Vikas Arora (vikasa at google dot com)
 - Vincent Rabaud (vrabaud at google dot com)
+- Vlad Tsyrklevich (vtsyrklevich at chromium dot org)
 - Yang Zhang (yang dot zhang at arm dot com)
+- Yannis Guyon (yguyon at google dot com)

Android.mk

@@ -85,11 +85,13 @@ dsp_dec_srcs := \
     src/dsp/upsampling_msa.c \
     src/dsp/upsampling_neon.$(NEON) \
     src/dsp/upsampling_sse2.c \
+    src/dsp/upsampling_sse41.c \
     src/dsp/yuv.c \
     src/dsp/yuv_mips32.c \
     src/dsp/yuv_mips_dsp_r2.c \
     src/dsp/yuv_neon.$(NEON) \
     src/dsp/yuv_sse2.c \
+    src/dsp/yuv_sse41.c \
 
 dsp_enc_srcs := \
     src/dsp/cost.c \
@@ -97,7 +99,6 @@ dsp_enc_srcs := \
     src/dsp/cost_mips_dsp_r2.c \
     src/dsp/cost_sse2.c \
     src/dsp/enc.c \
-    src/dsp/enc_avx2.c \
     src/dsp/enc_mips32.c \
     src/dsp/enc_mips_dsp_r2.c \
     src/dsp/enc_msa.c \
@@ -121,7 +122,6 @@ enc_srcs := \
     src/enc/backward_references_enc.c \
     src/enc/config_enc.c \
     src/enc/cost_enc.c \
-    src/enc/delta_palettization_enc.c \
     src/enc/filter_enc.c \
     src/enc/frame_enc.c \
     src/enc/histogram_enc.c \
@@ -173,7 +173,7 @@ LOCAL_SRC_FILES := \
     $(utils_dec_srcs) \
 
 LOCAL_CFLAGS := $(WEBP_CFLAGS)
-LOCAL_C_INCLUDES += $(LOCAL_PATH)/src
+LOCAL_EXPORT_C_INCLUDES += $(LOCAL_PATH)/src
 
 # prefer arm over thumb mode for performance gains
 LOCAL_ARM_MODE := arm
@@ -207,7 +207,7 @@ LOCAL_SRC_FILES := \
     $(utils_enc_srcs) \
 
 LOCAL_CFLAGS := $(WEBP_CFLAGS)
-LOCAL_C_INCLUDES += $(LOCAL_PATH)/src
+LOCAL_EXPORT_C_INCLUDES += $(LOCAL_PATH)/src
 
 # prefer arm over thumb mode for performance gains
 LOCAL_ARM_MODE := arm
@@ -230,7 +230,7 @@ include $(CLEAR_VARS)
 LOCAL_SRC_FILES := $(demux_srcs)
 
 LOCAL_CFLAGS := $(WEBP_CFLAGS)
-LOCAL_C_INCLUDES += $(LOCAL_PATH)/src
+LOCAL_EXPORT_C_INCLUDES += $(LOCAL_PATH)/src
 
 # prefer arm over thumb mode for performance gains
 LOCAL_ARM_MODE := arm
@@ -253,7 +253,7 @@ include $(CLEAR_VARS)
 LOCAL_SRC_FILES := $(mux_srcs)
 
 LOCAL_CFLAGS := $(WEBP_CFLAGS)
-LOCAL_C_INCLUDES += $(LOCAL_PATH)/src
+LOCAL_EXPORT_C_INCLUDES += $(LOCAL_PATH)/src
 
 # prefer arm over thumb mode for performance gains
 LOCAL_ARM_MODE := arm

CMakeLists.txt

@@ -1,81 +1,118 @@
-cmake_minimum_required(VERSION 2.8.7)
+cmake_minimum_required(VERSION 3.5)
 
-project(libwebp C)
+project(WebP C)
 
 # Options for coder / decoder executables.
 option(WEBP_ENABLE_SIMD "Enable any SIMD optimization." ON)
-option(WEBP_BUILD_CWEBP "Build the cwebp command line tool." OFF)
-option(WEBP_BUILD_DWEBP "Build the dwebp command line tool." OFF)
-option(WEBP_BUILD_GIF2WEBP "Build the gif2webp conversion tool." OFF)
-option(WEBP_BUILD_IMG2WEBP "Build the img2webp animation tool." OFF)
-option(WEBP_BUILD_WEBPINFO "Build the webpinfo command line tool." OFF)
+option(WEBP_BUILD_ANIM_UTILS "Build animation utilities." ON)
+option(WEBP_BUILD_CWEBP "Build the cwebp command line tool." ON)
+option(WEBP_BUILD_DWEBP "Build the dwebp command line tool." ON)
+option(WEBP_BUILD_GIF2WEBP "Build the gif2webp conversion tool." ON)
+option(WEBP_BUILD_IMG2WEBP "Build the img2webp animation tool." ON)
+option(WEBP_BUILD_VWEBP "Build the vwebp viewer tool." ON)
+option(WEBP_BUILD_WEBPINFO "Build the webpinfo command line tool." ON)
+option(WEBP_BUILD_WEBPMUX "Build the webpmux command line tool." ON)
+option(WEBP_BUILD_EXTRAS "Build extras." ON)
 option(WEBP_BUILD_WEBP_JS "Emscripten build of webp.js." OFF)
-option(WEBP_ENABLE_NEAR_LOSSLESS "Enable near-lossless encoding" ON)
-option(WEBP_EXPERIMENTAL_FEATURES "Build with experimental features." OFF)
-option(WEBP_ENABLE_SWAP_16BIT_CSP "Enable byte swap for 16 bit colorspaces." OFF)
+option(WEBP_NEAR_LOSSLESS "Enable near-lossless encoding" ON)
+option(WEBP_ENABLE_SWAP_16BIT_CSP "Enable byte swap for 16 bit colorspaces."
+       OFF)
 
 if(WEBP_BUILD_WEBP_JS)
   set(WEBP_ENABLE_SIMD OFF)
+  set(WEBP_BUILD_ANIM_UTILS OFF)
+  set(WEBP_BUILD_CWEBP OFF)
+  set(WEBP_BUILD_DWEBP OFF)
+  set(WEBP_BUILD_GIF2WEBP OFF)
+  set(WEBP_BUILD_IMG2WEBP OFF)
+  set(WEBP_BUILD_VWEBP OFF)
+  set(WEBP_BUILD_WEBPINFO OFF)
+  set(WEBP_BUILD_WEBPMUX OFF)
+  set(WEBP_BUILD_EXTRAS OFF)
 endif()
 
 set(WEBP_DEP_LIBRARIES)
 set(WEBP_DEP_INCLUDE_DIRS)
 
 if(NOT CMAKE_BUILD_TYPE)
-  set(CMAKE_BUILD_TYPE "Release" CACHE
-    "Build type: Release, Debug or RelWithDebInfo" STRING FORCE
-  )
+  set(CMAKE_BUILD_TYPE "Release"
+      CACHE "Build type: Release, Debug, MinSizeRel or RelWithDebInfo" STRING
+      FORCE)
 endif()
 
 # Include dependencies.
 include(cmake/deps.cmake)
 include(GNUInstallDirs)
 
-################################################################################
+# ##############################################################################
 # Options.
 if(WEBP_ENABLE_SWAP_16BIT_CSP)
   add_definitions(-DWEBP_SWAP_16BIT_CSP=1)
 endif()
 
-################################################################################
+set(prefix ${CMAKE_INSTALL_PREFIX})
+set(exec_prefix "\$\{prefix\}")
+set(libdir "\$\{prefix\}/lib")
+set(includedir "\$\{prefix\}/include")
+set(PTHREAD_LIBS ${CMAKE_THREAD_LIBS_INIT})
+
+# ##############################################################################
 # Android only.
 if(ANDROID)
   include_directories(${ANDROID_NDK}/sources/android/cpufeatures)
   add_library(cpufeatures STATIC
-    ${ANDROID_NDK}/sources/android/cpufeatures/cpu-features.c
-  )
+              ${ANDROID_NDK}/sources/android/cpufeatures/cpu-features.c)
   target_link_libraries(cpufeatures dl)
   set(WEBP_DEP_LIBRARIES ${WEBP_DEP_LIBRARIES} cpufeatures)
   set(WEBP_DEP_INCLUDE_DIRS ${WEBP_DEP_INCLUDE_DIRS}
-    ${ANDROID_NDK}/sources/android/cpufeatures
-  )
+      ${ANDROID_NDK}/sources/android/cpufeatures)
   add_definitions(-DHAVE_CPU_FEATURES_H=1)
   set(HAVE_CPU_FEATURES_H 1)
 else()
   set(HAVE_CPU_FEATURES_H 0)
 endif()
 
-################################################################################
-# WebP source files.
-# Read the Makefile.am to get the source files.
+function(configure_pkg_config FILE)
+  configure_file("${CMAKE_CURRENT_SOURCE_DIR}/${FILE}.in"
+                 "${CMAKE_CURRENT_BINARY_DIR}/${FILE}")
 
-# We expect the Makefiles to define the sources as defined in
-# the first regex. E.g.:
-# libimagedec_la_SOURCES  = image_dec.c image_dec.h
+  if(HAVE_MATH_LIBRARY)
+    # MSVC doesn't have libm
+    file(READ ${CMAKE_CURRENT_BINARY_DIR}/${FILE} data)
+    string(REPLACE "-lm" "" data ${data})
+    file(WRITE ${CMAKE_CURRENT_BINARY_DIR}/${FILE} ${data})
+  endif()
+
+  install(
+    FILES "${CMAKE_CURRENT_BINARY_DIR}/${FILE}"
+    DESTINATION ${CMAKE_INSTALL_LIBDIR}/pkgconfig
+  )
+endfunction()
+
+# ##############################################################################
+# WebP source files. Read the Makefile.am to get the source files.
+
+# We expect the Makefiles to define the sources as defined in the first regex.
+# E.g.: libimagedec_la_SOURCES  = image_dec.c image_dec.h
 function(parse_Makefile_am FOLDER VAR SRC_REGEX)
   file(READ ${FOLDER}/Makefile.am MAKEFILE_AM)
-  string(REGEX MATCHALL "${SRC_REGEX}_SOURCES[ ]*\\+?=[ ]+[0-9a-z\\._ ]*"
-    FILES_PER_LINE ${MAKEFILE_AM}
-  )
+  string(REGEX MATCHALL
+               "${SRC_REGEX}_SOURCES[ ]*\\+?=[ ]+[0-9a-z\\._ ]*"
+               FILES_PER_LINE
+               ${MAKEFILE_AM})
   set(SRCS ${${VAR}})
   foreach(FILES ${FILES_PER_LINE})
     string(FIND ${FILES} "=" OFFSET)
     math(EXPR OFFSET "${OFFSET} + 2")
-    string(SUBSTRING ${FILES} ${OFFSET} -1 FILES)
+    string(SUBSTRING ${FILES}
+                     ${OFFSET}
+                     -1
+                     FILES)
     if(FILES)
-      string(REGEX MATCHALL "[0-9a-z\\._]+"
-        FILES ${FILES}
-      )
+      string(REGEX MATCHALL
+                   "[0-9a-z\\._]+"
+                   FILES
+                   ${FILES})
       foreach(FILE ${FILES})
         list(APPEND SRCS ${FOLDER}/${FILE})
       endforeach()
@@ -85,16 +122,16 @@ function(parse_Makefile_am FOLDER VAR SRC_REGEX)
 endfunction()
 
 set(WEBP_SRC_DIR ${CMAKE_CURRENT_SOURCE_DIR}/src)
-parse_Makefile_am(${WEBP_SRC_DIR}/dec "WEBP_DEC_SRCS" "")
-parse_Makefile_am(${WEBP_SRC_DIR}/demux "WEBP_DEMUX_SRCS" "")
-parse_Makefile_am(${WEBP_SRC_DIR}/dsp "WEBP_DSP_COMMON_SRCS" "COMMON")
-parse_Makefile_am(${WEBP_SRC_DIR}/dsp "WEBP_DSP_ENC_SRCS" "ENC")
-parse_Makefile_am(${WEBP_SRC_DIR}/dsp "WEBP_DSP_ENC_SRCS" "dsp_[^ ]*")
-parse_Makefile_am(${WEBP_SRC_DIR}/dsp "WEBP_DSP_DEC_SRCS" "decode_[^ ]*")
-parse_Makefile_am(${WEBP_SRC_DIR}/enc "WEBP_ENC_SRCS" "")
-parse_Makefile_am(${WEBP_SRC_DIR}/utils "WEBP_UTILS_COMMON_SRCS" "COMMON")
-parse_Makefile_am(${WEBP_SRC_DIR}/utils "WEBP_UTILS_ENC_SRCS" "ENC")
-parse_Makefile_am(${WEBP_SRC_DIR}/utils "WEBP_UTILS_DEC_SRCS" "decode_[^ ]*")
+parse_makefile_am(${WEBP_SRC_DIR}/dec "WEBP_DEC_SRCS" "")
+parse_makefile_am(${WEBP_SRC_DIR}/demux "WEBP_DEMUX_SRCS" "")
+parse_makefile_am(${WEBP_SRC_DIR}/dsp "WEBP_DSP_COMMON_SRCS" "COMMON")
+parse_makefile_am(${WEBP_SRC_DIR}/dsp "WEBP_DSP_ENC_SRCS" "ENC")
+parse_makefile_am(${WEBP_SRC_DIR}/dsp "WEBP_DSP_ENC_SRCS" "dsp_[^ ]*")
+parse_makefile_am(${WEBP_SRC_DIR}/dsp "WEBP_DSP_DEC_SRCS" "decode_[^ ]*")
+parse_makefile_am(${WEBP_SRC_DIR}/enc "WEBP_ENC_SRCS" "")
+parse_makefile_am(${WEBP_SRC_DIR}/utils "WEBP_UTILS_COMMON_SRCS" "COMMON")
+parse_makefile_am(${WEBP_SRC_DIR}/utils "WEBP_UTILS_ENC_SRCS" "ENC")
+parse_makefile_am(${WEBP_SRC_DIR}/utils "WEBP_UTILS_DEC_SRCS" "decode_[^ ]*")
 
 # Remove the files specific to SIMD we don't use.
 foreach(FILE ${WEBP_SIMD_FILES_NOT_TO_INCLUDE})
@@ -102,7 +139,12 @@ foreach(FILE ${WEBP_SIMD_FILES_NOT_TO_INCLUDE})
   list(REMOVE_ITEM WEBP_DSP_DEC_SRCS ${FILE})
 endforeach()
 
-### Define the mandatory libraries.
+# Generate the config.h file.
+configure_file(${CMAKE_CURRENT_LIST_DIR}/cmake/config.h.in
+               ${CMAKE_CURRENT_BINARY_DIR}/src/webp/config.h)
+add_definitions(-DHAVE_CONFIG_H)
+
+# ##############################################################################
 # Build the webpdecoder library.
 if(MSVC)
   # avoid security warnings for e.g., fopen() used in the examples.
@@ -110,257 +152,455 @@ if(MSVC)
 else()
   add_definitions(-Wall)
 endif()
-include_directories(${CMAKE_CURRENT_SOURCE_DIR} ${WEBP_DEP_INCLUDE_DIRS})
+include_directories(${WEBP_DEP_INCLUDE_DIRS})
 add_library(webpdecode OBJECT ${WEBP_DEC_SRCS})
+target_include_directories(webpdecode
+                           PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
+                                   ${CMAKE_CURRENT_SOURCE_DIR})
 add_library(webpdspdecode OBJECT ${WEBP_DSP_COMMON_SRCS} ${WEBP_DSP_DEC_SRCS})
-add_library(webputilsdecode OBJECT ${WEBP_UTILS_COMMON_SRCS}
-  ${WEBP_UTILS_DEC_SRCS})
-add_library(webpdecoder $<TARGET_OBJECTS:webpdecode>
-  $<TARGET_OBJECTS:webpdspdecode> $<TARGET_OBJECTS:webputilsdecode>)
+target_include_directories(webpdspdecode
+                           PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
+                                   ${CMAKE_CURRENT_SOURCE_DIR})
+add_library(webputilsdecode
+            OBJECT
+            ${WEBP_UTILS_COMMON_SRCS}
+            ${WEBP_UTILS_DEC_SRCS})
+target_include_directories(webputilsdecode
+                           PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
+                                   ${CMAKE_CURRENT_SOURCE_DIR})
+add_library(webpdecoder
+            $<TARGET_OBJECTS:webpdecode>
+            $<TARGET_OBJECTS:webpdspdecode>
+            $<TARGET_OBJECTS:webputilsdecode>)
 target_link_libraries(webpdecoder ${WEBP_DEP_LIBRARIES})
+target_include_directories(
+  webpdecoder
+  PRIVATE ${CMAKE_CURRENT_BINARY_DIR} ${CMAKE_CURRENT_SOURCE_DIR}
+  INTERFACE $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}>
+            $<INSTALL_INTERFACE:${CMAKE_INSTALL_INCLUDEDIR}>)
+set_target_properties(
+  webpdecoder
+  PROPERTIES PUBLIC_HEADER "${CMAKE_CURRENT_SOURCE_DIR}/src/webp/decode.h;\
+${CMAKE_CURRENT_SOURCE_DIR}/src/webp/types.h")
+
+configure_pkg_config("src/libwebpdecoder.pc")
 
 # Build the webp library.
 add_library(webpencode OBJECT ${WEBP_ENC_SRCS})
-target_include_directories(
-  webpencode PRIVATE ${CMAKE_CURRENT_BINARY_DIR} ${CMAKE_CURRENT_SOURCE_DIR}
-                     ${CMAKE_CURRENT_SOURCE_DIR}/src)
-add_library(webpdsp OBJECT ${WEBP_DSP_COMMON_SRCS} ${WEBP_DSP_DEC_SRCS}
-                           ${WEBP_DSP_ENC_SRCS})
-target_include_directories(webpdsp PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
-                                           ${CMAKE_CURRENT_SOURCE_DIR})
-add_library(webputils OBJECT ${WEBP_UTILS_COMMON_SRCS} ${WEBP_UTILS_DEC_SRCS}
-                             ${WEBP_UTILS_ENC_SRCS})
-target_include_directories(webputils PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
-                                             ${CMAKE_CURRENT_SOURCE_DIR})
-add_library(webp $<TARGET_OBJECTS:webpdecode> $<TARGET_OBJECTS:webpdsp>
-                 $<TARGET_OBJECTS:webpencode> $<TARGET_OBJECTS:webputils>)
-if(XCODE)
-  libwebp_add_stub_file(webp)
-endif()
+target_include_directories(webpencode
+                           PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
+                                   ${CMAKE_CURRENT_SOURCE_DIR})
+add_library(webpdsp
+            OBJECT
+            ${WEBP_DSP_COMMON_SRCS}
+            ${WEBP_DSP_DEC_SRCS}
+            ${WEBP_DSP_ENC_SRCS})
+target_include_directories(webpdsp
+                           PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
+                                   ${CMAKE_CURRENT_SOURCE_DIR})
+add_library(webputils
+            OBJECT
+            ${WEBP_UTILS_COMMON_SRCS}
+            ${WEBP_UTILS_DEC_SRCS}
+            ${WEBP_UTILS_ENC_SRCS})
+target_include_directories(webputils
+                           PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
+                                   ${CMAKE_CURRENT_SOURCE_DIR})
+add_library(webp
+            $<TARGET_OBJECTS:webpdecode>
+            $<TARGET_OBJECTS:webpdsp>
+            $<TARGET_OBJECTS:webpencode>
+            $<TARGET_OBJECTS:webputils>)
 target_link_libraries(webp ${WEBP_DEP_LIBRARIES})
-target_include_directories(
-  webp PRIVATE ${CMAKE_CURRENT_SOURCE_DIR} ${CMAKE_CURRENT_BINARY_DIR}
-  PUBLIC $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/src>
-         $<INSTALL_INTERFACE:include>)
+target_include_directories(webp
+                           PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}
+                                   ${CMAKE_CURRENT_BINARY_DIR}
+                           PUBLIC $<INSTALL_INTERFACE:include>)
 set_target_properties(
   webp
   PROPERTIES PUBLIC_HEADER "${CMAKE_CURRENT_SOURCE_DIR}/src/webp/decode.h;\
 ${CMAKE_CURRENT_SOURCE_DIR}/src/webp/encode.h;\
 ${CMAKE_CURRENT_SOURCE_DIR}/src/webp/types.h")
 
-# Make sure the OBJECT libraries are built with position independent code
-# (it is not ON by default).
-set_target_properties(webpdecode webpdspdecode webputilsdecode
-  webpencode webpdsp webputils PROPERTIES POSITION_INDEPENDENT_CODE ON)
+# Make sure the OBJECT libraries are built with position independent code (it is
+# not ON by default).
+set_target_properties(webpdecode
+                      webpdspdecode
+                      webputilsdecode
+                      webpencode
+                      webpdsp
+                      webputils
+                      PROPERTIES POSITION_INDEPENDENT_CODE ON)
+configure_pkg_config("src/libwebp.pc")
 
 # Build the webp demux library.
 add_library(webpdemux ${WEBP_DEMUX_SRCS})
 target_link_libraries(webpdemux webp)
-target_include_directories(
-  webpdemux PRIVATE ${CMAKE_CURRENT_SOURCE_DIR} ${CMAKE_CURRENT_BINARY_DIR}
-  PUBLIC $<INSTALL_INTERFACE:include>)
+target_include_directories(webpdemux
+                           PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}
+                                   ${CMAKE_CURRENT_BINARY_DIR}
+                           PUBLIC $<INSTALL_INTERFACE:include>)
 set_target_properties(
   webpdemux
-  PROPERTIES
-    PUBLIC_HEADER
-    "${CMAKE_CURRENT_SOURCE_DIR}/src/webp/decode.h;\
+  PROPERTIES PUBLIC_HEADER "${CMAKE_CURRENT_SOURCE_DIR}/src/webp/decode.h;\
 ${CMAKE_CURRENT_SOURCE_DIR}/src/webp/demux.h;\
 ${CMAKE_CURRENT_SOURCE_DIR}/src/webp/mux_types.h;\
 ${CMAKE_CURRENT_SOURCE_DIR}/src/webp/types.h")
 
+configure_pkg_config("src/demux/libwebpdemux.pc")
+
 # Set the version numbers.
 function(parse_version FILE NAME VAR)
   file(READ ${CMAKE_CURRENT_SOURCE_DIR}/src/${FILE} SOURCE_FILE)
-  string(REGEX MATCH "${NAME}_la_LDFLAGS[^\n]* -version-info [0-9:]+" TMP
-    ${SOURCE_FILE})
-  string(REGEX MATCH "[0-9:]+" TMP ${TMP})
-  string(REGEX REPLACE ":" "." VERSION ${TMP})
+  string(REGEX MATCH
+               "${NAME}_la_LDFLAGS[^\n]* -version-info [0-9:]+"
+               TMP
+               ${SOURCE_FILE})
+  string(REGEX MATCH
+               "[0-9:]+"
+               TMP
+               ${TMP})
+  string(REGEX
+         REPLACE ":"
+                 "."
+                 VERSION
+                 ${TMP})
   set(${VAR} "${VERSION}" PARENT_SCOPE)
 endfunction()
 parse_version(Makefile.am webp WEBP_WEBP_SOVERSION)
-set_target_properties(webp PROPERTIES VERSION ${PACKAGE_VERSION}
-  SOVERSION ${WEBP_WEBP_SOVERSION})
+set_target_properties(webp
+                      PROPERTIES VERSION
+                                 ${PACKAGE_VERSION}
+                                 SOVERSION
+                                 ${WEBP_WEBP_SOVERSION})
 parse_version(Makefile.am webpdecoder WEBP_DECODER_SOVERSION)
-set_target_properties(webpdecoder PROPERTIES VERSION ${PACKAGE_VERSION}
-  SOVERSION ${WEBP_DECODER_SOVERSION})
+set_target_properties(webpdecoder
+                      PROPERTIES VERSION
+                                 ${PACKAGE_VERSION}
+                                 SOVERSION
+                                 ${WEBP_DECODER_SOVERSION})
 parse_version(demux/Makefile.am webpdemux WEBP_DEMUX_SOVERSION)
-set_target_properties(webpdemux PROPERTIES VERSION ${PACKAGE_VERSION}
-  SOVERSION ${WEBP_DEMUX_SOVERSION})
+set_target_properties(webpdemux
+                      PROPERTIES VERSION
+                                 ${PACKAGE_VERSION}
+                                 SOVERSION
+                                 ${WEBP_DEMUX_SOVERSION})
+file(READ ${CMAKE_CURRENT_SOURCE_DIR}/configure.ac CONFIGURE_FILE)
+string(REGEX MATCH
+             "AC_INIT\\([^\n]*\\[[0-9\\.]+\\]"
+             TMP
+             ${CONFIGURE_FILE})
+string(REGEX MATCH
+             "[0-9\\.]+"
+             PROJECT_VERSION
+             ${TMP})
 
 # Define the libraries to install.
 set(INSTALLED_LIBRARIES webpdecoder webp webpdemux)
 
-### Deal with SIMD.
-# Change the compile flags for SIMD files we use.
+# Deal with SIMD. Change the compile flags for SIMD files we use.
 list(LENGTH WEBP_SIMD_FILES_TO_INCLUDE WEBP_SIMD_FILES_TO_INCLUDE_LENGTH)
 math(EXPR WEBP_SIMD_FILES_TO_INCLUDE_RANGE
-  "${WEBP_SIMD_FILES_TO_INCLUDE_LENGTH}-1"
-)
+     "${WEBP_SIMD_FILES_TO_INCLUDE_LENGTH}-1")
 
 foreach(I_FILE RANGE ${WEBP_SIMD_FILES_TO_INCLUDE_RANGE})
   list(GET WEBP_SIMD_FILES_TO_INCLUDE ${I_FILE} FILE)
   list(GET WEBP_SIMD_FLAGS_TO_INCLUDE ${I_FILE} SIMD_COMPILE_FLAG)
-  set_source_files_properties(${FILE} PROPERTIES
-    COMPILE_FLAGS ${SIMD_COMPILE_FLAG}
-  )
+  set_source_files_properties(${FILE}
+                              PROPERTIES
+                              COMPILE_FLAGS
+                              ${SIMD_COMPILE_FLAG})
 endforeach()
 
-# Build the executables if asked for.
-if(WEBP_BUILD_CWEBP OR WEBP_BUILD_DWEBP OR
-   WEBP_BUILD_GIF2WEBP OR WEBP_BUILD_IMG2WEBP OR WEBP_BUILD_WEBP_JS)
-  # Example utility library.
-  parse_Makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "EXAMPLEUTIL_SRCS"
-    "example_util_[^ ]*")
-  list(APPEND EXAMPLEUTIL_SRCS
-    ${CMAKE_CURRENT_SOURCE_DIR}/examples/stopwatch.h)
-  add_library(exampleutil ${EXAMPLEUTIL_SRCS})
+if(WEBP_BUILD_GIF2WEBP AND NOT GIF_FOUND)
+  set(WEBP_BUILD_GIF2WEBP OFF)
+endif()
 
-  parse_Makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/imageio "IMAGEIOUTILS_SRCS"
-    "imageio_util_[^ ]*")
+if(WEBP_BUILD_ANIM_UTILS AND NOT GIF_FOUND)
+  set(WEBP_BUILD_ANIM_UTILS OFF)
+endif()
+
+# Build the executables if asked for.
+if(WEBP_BUILD_ANIM_UTILS
+   OR WEBP_BUILD_CWEBP
+   OR WEBP_BUILD_DWEBP
+   OR WEBP_BUILD_GIF2WEBP
+   OR WEBP_BUILD_IMG2WEBP
+   OR WEBP_BUILD_VWEBP)
+  # Example utility library.
+  parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "EXAMPLEUTIL_SRCS"
+                    "example_util_[^ ]*")
+  list(APPEND EXAMPLEUTIL_SRCS ${CMAKE_CURRENT_SOURCE_DIR}/examples/stopwatch.h)
+  add_library(exampleutil ${EXAMPLEUTIL_SRCS})
+  target_include_directories(
+    exampleutil
+    PUBLIC $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/src>)
+
+  parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/imageio "IMAGEIOUTILS_SRCS"
+                    "imageio_util_[^ ]*")
   add_library(imageioutil ${IMAGEIOUTILS_SRCS})
   target_link_libraries(imageioutil webp)
 
   # Image-decoding utility library.
-  parse_Makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/imageio "IMAGEDEC_SRCS"
-    "imagedec_[^ ]*")
+  parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/imageio "IMAGEDEC_SRCS"
+                    "imagedec_[^ ]*")
   add_library(imagedec ${IMAGEDEC_SRCS})
-  target_link_libraries(imagedec imageioutil webp ${WEBP_DEP_IMG_LIBRARIES})
+  target_link_libraries(imagedec
+                        imageioutil
+                        webpdemux
+                        webp
+                        ${WEBP_DEP_IMG_LIBRARIES})
 
   # Image-encoding utility library.
-  parse_Makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/imageio "IMAGEENC_SRCS"
-    "imageenc_[^ ]*")
+  parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/imageio "IMAGEENC_SRCS"
+                    "imageenc_[^ ]*")
   add_library(imageenc ${IMAGEENC_SRCS})
   target_link_libraries(imageenc webp)
 
-  set_property(TARGET exampleutil imageioutil imagedec imageenc
-    PROPERTY INCLUDE_DIRECTORIES
-    ${CMAKE_CURRENT_SOURCE_DIR}/src ${CMAKE_CURRENT_BINARY_DIR}/src)
+  set_property(TARGET exampleutil
+                      imageioutil
+                      imagedec
+                      imageenc
+               PROPERTY INCLUDE_DIRECTORIES ${CMAKE_CURRENT_SOURCE_DIR}/src
+                        ${CMAKE_CURRENT_BINARY_DIR}/src)
 endif()
 
 if(WEBP_BUILD_DWEBP)
   # dwebp
-  include_directories(${WEBP_DEP_IMG_INCLUDE_DIRS})
-  parse_Makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "DWEBP_SRCS"
-    "dwebp")
+  parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "DWEBP_SRCS" "dwebp")
   add_executable(dwebp ${DWEBP_SRCS})
-  target_link_libraries(dwebp exampleutil imagedec imageenc webpdecoder)
-  install(TARGETS dwebp RUNTIME DESTINATION bin)
-  set_property(TARGET dwebp PROPERTY INCLUDE_DIRECTORIES
-    ${CMAKE_CURRENT_SOURCE_DIR}/src ${CMAKE_CURRENT_BINARY_DIR}/src)
+  target_link_libraries(dwebp exampleutil imagedec imageenc)
+  target_include_directories(dwebp PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/src)
+  install(TARGETS dwebp RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
 endif()
 
 if(WEBP_BUILD_CWEBP)
   # cwebp
-  include_directories(${WEBP_DEP_IMG_INCLUDE_DIRS})
-  parse_Makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "CWEBP_SRCS"
-    "cwebp")
+  parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "CWEBP_SRCS" "cwebp")
   add_executable(cwebp ${CWEBP_SRCS})
   target_link_libraries(cwebp exampleutil imagedec webp)
-  install(TARGETS cwebp RUNTIME DESTINATION bin)
-  set_property(TARGET cwebp PROPERTY INCLUDE_DIRECTORIES
-    ${CMAKE_CURRENT_SOURCE_DIR}/src ${CMAKE_CURRENT_BINARY_DIR}/src)
-endif()
-
-if(WEBP_BUILD_GIF2WEBP AND NOT GIF_FOUND)
-  unset(WEBP_BUILD_GIF2WEBP CACHE)
+  target_include_directories(cwebp PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/src)
+  install(TARGETS cwebp RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
 endif()
 
 if(WEBP_BUILD_GIF2WEBP OR WEBP_BUILD_IMG2WEBP)
-  parse_Makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/src/mux "WEBP_MUX_SRCS"
-    "")
-  add_library(webpmux ${WEBP_MUX_SRCS})
-  target_link_libraries(webpmux webp)
+  parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/src/mux "WEBP_MUX_SRCS" "")
+  add_library(libwebpmux ${WEBP_MUX_SRCS})
+  target_link_libraries(libwebpmux webp)
+  target_include_directories(libwebpmux
+                             PRIVATE ${CMAKE_CURRENT_BINARY_DIR}
+                                     ${CMAKE_CURRENT_SOURCE_DIR})
   parse_version(mux/Makefile.am webpmux WEBP_MUX_SOVERSION)
-  set_target_properties(webpmux PROPERTIES VERSION ${PACKAGE_VERSION}
-    SOVERSION ${WEBP_MUX_SOVERSION})
-  set_target_properties(
-    webpmux
-    PROPERTIES PUBLIC_HEADER "${CMAKE_CURRENT_SOURCE_DIR}/src/webp/mux.h;\
+  set_target_properties(libwebpmux
+                        PROPERTIES VERSION
+                                   ${PACKAGE_VERSION}
+                                   SOVERSION
+                                   ${WEBP_MUX_SOVERSION})
+  set_target_properties(libwebpmux
+                        PROPERTIES PUBLIC_HEADER
+                                   "${CMAKE_CURRENT_SOURCE_DIR}/src/webp/mux.h;\
 ${CMAKE_CURRENT_SOURCE_DIR}/src/webp/mux_types.h;\
 ${CMAKE_CURRENT_SOURCE_DIR}/src/webp/types.h;")
-  list(APPEND INSTALLED_LIBRARIES webpmux)
+  set_target_properties(libwebpmux PROPERTIES OUTPUT_NAME webpmux)
+  list(APPEND INSTALLED_LIBRARIES libwebpmux)
+  configure_pkg_config("src/mux/libwebpmux.pc")
 endif()
 
 if(WEBP_BUILD_GIF2WEBP)
   # gif2webp
   include_directories(${WEBP_DEP_GIF_INCLUDE_DIRS})
-  parse_Makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "GIF2WEBP_SRCS"
-    "gif2webp")
+  parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "GIF2WEBP_SRCS"
+                    "gif2webp")
   add_executable(gif2webp ${GIF2WEBP_SRCS})
-  target_link_libraries(gif2webp exampleutil imageioutil webp webpmux
-    ${WEBP_DEP_GIF_LIBRARIES})
-  install(TARGETS gif2webp RUNTIME DESTINATION bin)
-  set_property(TARGET gif2webp PROPERTY INCLUDE_DIRECTORIES
-    ${CMAKE_CURRENT_SOURCE_DIR}/src ${CMAKE_CURRENT_BINARY_DIR}/src)
+  target_link_libraries(gif2webp
+                        exampleutil
+                        imageioutil
+                        webp
+                        libwebpmux
+                        ${WEBP_DEP_GIF_LIBRARIES})
+  target_include_directories(gif2webp PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/src)
+  install(TARGETS gif2webp RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
 endif()
 
 if(WEBP_BUILD_IMG2WEBP)
   # img2webp
   include_directories(${WEBP_DEP_IMG_INCLUDE_DIRS})
-  parse_Makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "IMG2WEBP_SRCS"
-    "img2webp")
+  parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "IMG2WEBP_SRCS"
+                    "img2webp")
   add_executable(img2webp ${IMG2WEBP_SRCS})
-  target_link_libraries(img2webp exampleutil imagedec imageioutil webp webpmux)
-  install(TARGETS img2webp RUNTIME DESTINATION bin)
-  set_property(TARGET img2webp PROPERTY INCLUDE_DIRECTORIES
-    ${CMAKE_CURRENT_SOURCE_DIR}/src ${CMAKE_CURRENT_BINARY_DIR}/src)
+  target_link_libraries(img2webp
+                        exampleutil
+                        imagedec
+                        imageioutil
+                        webp
+                        libwebpmux)
+  target_include_directories(img2webp PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/src)
+  install(TARGETS img2webp RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
 endif()
 
-if (WEBP_BUILD_WEBPINFO)
+if(WEBP_BUILD_VWEBP)
+  # vwebp
+  find_package(GLUT)
+  if(GLUT_FOUND)
+    include_directories(${WEBP_DEP_IMG_INCLUDE_DIRS})
+    parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "VWEBP_SRCS" "vwebp")
+    add_executable(vwebp ${VWEBP_SRCS})
+    target_link_libraries(vwebp
+                          ${OPENGL_LIBRARIES}
+                          exampleutil
+                          GLUT::GLUT
+                          imageioutil
+                          webp
+                          webpdemux)
+    target_include_directories(vwebp
+                               PRIVATE GLUT::GLUT
+                                       ${CMAKE_CURRENT_BINARY_DIR}/src
+                                       ${OPENGL_INCLUDE_DIR})
+    install(TARGETS vwebp RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
+    if(${CMAKE_SYSTEM_NAME} STREQUAL "Darwin")
+      check_c_compiler_flag("-Wno-deprecated-declarations" HAS_NO_DEPRECATED)
+      if(HAS_NO_DEPRECATED)
+        target_compile_options(vwebp PRIVATE "-Wno-deprecated-declarations")
+      endif()
+    endif()
+  endif()
+endif()
+
+if(WEBP_BUILD_WEBPINFO)
   # webpinfo
   include_directories(${WEBP_DEP_IMG_INCLUDE_DIRS})
-  parse_Makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "WEBPINFO_SRCS"
-    "webpinfo")
+  parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "WEBPINFO_SRCS"
+                    "webpinfo")
   add_executable(webpinfo ${WEBPINFO_SRCS})
   target_link_libraries(webpinfo exampleutil imageioutil)
-  install(TARGETS webpinfo RUNTIME DESTINATION bin)
-  set_property(TARGET webpinfo PROPERTY INCLUDE_DIRECTORIES
-    ${CMAKE_CURRENT_SOURCE_DIR}/src ${CMAKE_CURRENT_BINARY_DIR}/src)
+  target_include_directories(webpinfo PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/src)
+  install(TARGETS webpinfo RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
+endif()
+
+if(WEBP_BUILD_WEBPMUX)
+  # webpmux
+  parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "WEBPMUX_SRCS"
+                    "webpmux")
+  add_executable(webpmux ${WEBPMUX_SRCS})
+  target_link_libraries(webpmux exampleutil imageioutil libwebpmux webp)
+  target_include_directories(webpmux PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/src)
+  install(TARGETS webpmux RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
+endif()
+
+if(WEBP_BUILD_EXTRAS)
+  set(EXTRAS_MAKEFILE "${CMAKE_CURRENT_SOURCE_DIR}/extras")
+  parse_makefile_am(${EXTRAS_MAKEFILE} "WEBP_EXTRAS_SRCS" "libwebpextras_la")
+  parse_makefile_am(${EXTRAS_MAKEFILE} "GET_DISTO_SRCS" "get_disto")
+  parse_makefile_am(${EXTRAS_MAKEFILE} "WEBP_QUALITY_SRCS" "webp_quality")
+  parse_makefile_am(${EXTRAS_MAKEFILE} "VWEBP_SDL_SRCS" "vwebp_sdl")
+
+  # get_disto
+  add_executable(get_disto ${GET_DISTO_SRCS})
+  target_link_libraries(get_disto imagedec)
+  target_include_directories(get_disto
+                             PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}
+                                     ${CMAKE_CURRENT_SOURCE_DIR}/src
+                                     ${CMAKE_CURRENT_BINARY_DIR}/src)
+  install(TARGETS get_disto RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
+
+  # webp_quality
+  add_executable(webp_quality ${WEBP_QUALITY_SRCS} ${WEBP_EXTRAS_SRCS})
+  target_link_libraries(webp_quality exampleutil imagedec)
+  target_include_directories(webp_quality PRIVATE ${CMAKE_CURRENT_SOURCE_DIR})
+  install(TARGETS webp_quality RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
+
+  # vwebp_sdl
+  find_package(SDL)
+  if(SDL_FOUND)
+    add_executable(vwebp_sdl ${VWEBP_SDL_SRCS})
+    target_link_libraries(vwebp_sdl ${SDL_LIBRARY} imageioutil webp)
+    target_include_directories(vwebp_sdl
+                               PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}
+                                       ${CMAKE_CURRENT_SOURCE_DIR}/src
+                                       ${CMAKE_CURRENT_BINARY_DIR}
+                                       ${CMAKE_CURRENT_BINARY_DIR}/src
+                                       ${SDL_INCLUDE_DIR})
+    set(WEBP_HAVE_SDL 1)
+    target_compile_definitions(vwebp_sdl PUBLIC WEBP_HAVE_SDL)
+    install(TARGETS vwebp_sdl RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
+  endif()
 endif()
 
 if(WEBP_BUILD_WEBP_JS)
   # JavaScript version
-  add_executable(webp_js
-                 ${CMAKE_CURRENT_SOURCE_DIR}/extras/webp_to_sdl.c)
+  add_executable(webp_js ${CMAKE_CURRENT_SOURCE_DIR}/extras/webp_to_sdl.c)
   target_link_libraries(webp_js webpdecoder SDL)
+  target_include_directories(webp_js PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
   set(WEBP_HAVE_SDL 1)
-  set_target_properties(webp_js PROPERTIES LINK_FLAGS
-      "-s EXPORTED_FUNCTIONS='[\"_WebpToSDL\"]' -s INVOKE_RUN=0")
+  set_target_properties(
+    webp_js
+    PROPERTIES LINK_FLAGS
+               "-s EXPORTED_FUNCTIONS='[\"_WebpToSDL\"]' -s INVOKE_RUN=0 \
+       -s EXTRA_EXPORTED_RUNTIME_METHODS='[\"cwrap\"]'")
   set_target_properties(webp_js PROPERTIES OUTPUT_NAME webp)
   target_compile_definitions(webp_js PUBLIC EMSCRIPTEN WEBP_HAVE_SDL)
 
   # WASM version
-  add_executable(webp_wasm
-                 ${CMAKE_CURRENT_SOURCE_DIR}/extras/webp_to_sdl.c)
+  add_executable(webp_wasm ${CMAKE_CURRENT_SOURCE_DIR}/extras/webp_to_sdl.c)
   target_link_libraries(webp_wasm webpdecoder SDL)
-  set_target_properties(webp_wasm PROPERTIES LINK_FLAGS
-      "-s WASM=1 -s 'BINARYEN_METHOD=\"native-wasm\"' \
-      -s EXPORTED_FUNCTIONS='[\"_WebpToSDL\"]' -s INVOKE_RUN=0")
+  target_include_directories(webp_wasm PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
+  set_target_properties(
+    webp_wasm
+    PROPERTIES LINK_FLAGS "-s WASM=1 -s 'BINARYEN_METHOD=\"native-wasm\"' \
+       -s EXPORTED_FUNCTIONS='[\"_WebpToSDL\"]' -s INVOKE_RUN=0 \
+       -s EXTRA_EXPORTED_RUNTIME_METHODS='[\"cwrap\"]'")
   target_compile_definitions(webp_wasm PUBLIC EMSCRIPTEN WEBP_HAVE_SDL)
 
-  target_compile_definitions(webpdecoder PUBLIC EMSCRIPTEN)
+  target_compile_definitions(webpdspdecode PUBLIC EMSCRIPTEN)
 endif()
 
-# Generate the config.h file.
-configure_file(${CMAKE_CURRENT_LIST_DIR}/cmake/config.h.in
-  ${CMAKE_CURRENT_BINARY_DIR}/src/webp/config.h)
-add_definitions(-DHAVE_CONFIG_H)
-# The webp folder is included as we reference config.h as
-# ../webp/config.h or webp/config.h
-include_directories(${CMAKE_CURRENT_BINARY_DIR})
+if(WEBP_BUILD_ANIM_UTILS)
+  # anim_diff
+  include_directories(${WEBP_DEP_IMG_INCLUDE_DIRS} ${WEBP_DEP_GIF_INCLUDE_DIRS})
+  parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "ANIM_DIFF_SRCS"
+                    "anim_diff")
+  add_executable(anim_diff ${ANIM_DIFF_SRCS})
+  target_link_libraries(anim_diff
+                        exampleutil
+                        imagedec
+                        imageenc
+                        imageioutil
+                        webp
+                        libwebpmux
+                        webpdemux
+                        ${WEBP_DEP_GIF_LIBRARIES})
+  target_include_directories(anim_diff PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/src)
+
+  # anim_dump
+  include_directories(${WEBP_DEP_IMG_INCLUDE_DIRS} ${WEBP_DEP_GIF_INCLUDE_DIRS})
+  parse_makefile_am(${CMAKE_CURRENT_SOURCE_DIR}/examples "ANIM_DUMP_SRCS"
+                    "anim_dump")
+  add_executable(anim_dump ${ANIM_DUMP_SRCS})
+  target_link_libraries(anim_dump
+                        exampleutil
+                        imagedec
+                        imageenc
+                        imageioutil
+                        webp
+                        libwebpmux
+                        webpdemux
+                        ${WEBP_DEP_GIF_LIBRARIES})
+  target_include_directories(anim_dump PRIVATE ${CMAKE_CURRENT_BINARY_DIR}/src)
+endif()
 
 # Install the different headers and libraries.
-install(
-  TARGETS ${INSTALLED_LIBRARIES}
-  EXPORT WebPTargets
-  PUBLIC_HEADER DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/webp
-  INCLUDES
-  DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}
-  ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR}
-  LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
-  RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
-set(ConfigPackageLocation ${CMAKE_INSTALL_DATADIR}/WebP/cmake/)
-install(EXPORT WebPTargets NAMESPACE WebP::
+include(GNUInstallDirs)
+install(TARGETS ${INSTALLED_LIBRARIES}
+        EXPORT ${PROJECT_NAME}Targets
+        PUBLIC_HEADER DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}/webp
+        INCLUDES
+        DESTINATION ${CMAKE_INSTALL_INCLUDEDIR}
+        ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR}
+        LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
+        RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
+set(ConfigPackageLocation ${CMAKE_INSTALL_DATADIR}/${PROJECT_NAME}/cmake/)
+install(EXPORT ${PROJECT_NAME}Targets
+        NAMESPACE ${PROJECT_NAME}::
         DESTINATION ${ConfigPackageLocation})
 
 # Create the CMake version file.
@@ -368,30 +608,38 @@ include(CMakePackageConfigHelpers)
 write_basic_package_version_file(
   "${CMAKE_CURRENT_BINARY_DIR}/WebPConfigVersion.cmake"
   VERSION ${PACKAGE_VERSION}
-  COMPATIBILITY AnyNewerVersion
-)
+  COMPATIBILITY AnyNewerVersion)
 
 # Create the Config file.
 include(CMakePackageConfigHelpers)
-set(ConfigPackageLocation share/WebP/cmake/)
 configure_package_config_file(
   ${CMAKE_CURRENT_SOURCE_DIR}/cmake/WebPConfig.cmake.in
   ${CMAKE_CURRENT_BINARY_DIR}/WebPConfig.cmake
-  INSTALL_DESTINATION ${ConfigPackageLocation}
-  PATH_VARS CMAKE_INSTALL_INCLUDEDIR)
+  INSTALL_DESTINATION
+  ${ConfigPackageLocation})
 
 # Install the generated CMake files.
-install(
-  FILES "${CMAKE_CURRENT_BINARY_DIR}/WebPConfigVersion.cmake"
-        "${CMAKE_CURRENT_BINARY_DIR}/WebPConfig.cmake"
-  DESTINATION ${ConfigPackageLocation}
-)
+install(FILES "${CMAKE_CURRENT_BINARY_DIR}/WebPConfigVersion.cmake"
+              "${CMAKE_CURRENT_BINARY_DIR}/WebPConfig.cmake"
+        DESTINATION ${ConfigPackageLocation})
 
 # Install the man pages.
-set(MAN_PAGES cwebp.1 dwebp.1 gif2webp.1 img2webp.1 vwebp.1 webpmux.1
-  webpinfo.1)
-set(EXEC_BUILDS "CWEBP" "DWEBP" "GIF2WEBP" "IMG2WEBP" "VWEBP" "WEBPMUX"
-  "WEBPINFO")
+set(MAN_PAGES
+    cwebp.1
+    dwebp.1
+    gif2webp.1
+    img2webp.1
+    vwebp.1
+    webpmux.1
+    webpinfo.1)
+set(EXEC_BUILDS
+    "CWEBP"
+    "DWEBP"
+    "GIF2WEBP"
+    "IMG2WEBP"
+    "VWEBP"
+    "WEBPMUX"
+    "WEBPINFO")
 list(LENGTH MAN_PAGES MAN_PAGES_LENGTH)
 math(EXPR MAN_PAGES_RANGE "${MAN_PAGES_LENGTH} - 1")
 
@@ -400,8 +648,7 @@ foreach(I_MAN RANGE ${MAN_PAGES_RANGE})
   if(WEBP_BUILD_${EXEC_BUILD})
     list(GET MAN_PAGES ${I_MAN} MAN_PAGE)
     install(FILES ${CMAKE_CURRENT_SOURCE_DIR}/man/${MAN_PAGE}
-      DESTINATION ${CMAKE_INSTALL_PREFIX}/share/man/man1
-      COMPONENT doc
-    )
+            DESTINATION ${CMAKE_INSTALL_MANDIR}/man1
+            COMPONENT doc)
   endif()
 endforeach()

ChangeLog

@@ -1,9 +1,185 @@
+fa8210e4 Fix pair update in stochastic entropy merging.
+825389ac README.mux: add a reference to the AnimDecoder API
+3be698c3 CMake: fix webp_js compilation
+4cbb4caf update NEWS
+f5a5918d bump version to 1.0.1
+d61385db Speed-up: Make sure we only initialize histograms when needed.
+0c570316 update AUTHORS
+301a2dda img2webp: add help note about arguments from a file
+f0abab92 Speedups for empty histograms.
+f2dfd925 Split HistogramAdd to only have the high level logic in C.
+06b7bc7d Fix compilation on windows and clang-cl+ninja.
+98179495 remove some minor TODOs
+cbf82cc0 Remove AVX2 files.
+5030e902 Merge "TIFF decoder: remove unused KINV definition"
+ac543311 Remove a few more useless #defines
+123d3306 TIFF decoder: remove unused KINV definition
+ef1094b0 Merge "- install pkg-config files during the CMake build"
+b911fbc9 libwebp: Remove duplicate GIFDisplayError in anim_util
+eee00b66 - install pkg-config files during the CMake build
+ac3ec8c9 Merge "Clean-up the common sources in dsp."
+3e13da7b Clean-up the common sources in dsp.
+5c395f1d libwebp: cmake-format all
+e7a69729 libwebp: Add extras targets in CMakeLists.txt
+e52485d6 libwebp: Rename macros in webpmux.c
+92dc0f09 clean-up MakeInputImageCopy()
+39952de2 VP8IteratorImport: add missing 'const'
+382af7a2 clean-up WebPBlendAlpha
+14d020f6 libwebp: Use ExUtilGet*() in anim_diff
+0d92ff25 libwebp: remove useless variable in gif2webp
+556cb1b4 Merge "CMake: Set WEBP_BUILD_GIF2WEBP to off"
+da26ee49 CMake: Set WEBP_BUILD_GIF2WEBP to off
+b2a867c0 cwebp: Don't premultiply during -resize if -exact
+637141bc pngdec: fix build w/libpng < 1.4.x
+bc5092b1 pngdec: set memory functions
+50d8345a Fix CMake math library.
+6aa3e8aa Fix math library on Visual Studio.
+d71df4e2 Fix math library finding in CMake.
+de08d727 cosmetics: normalize include guard comment
+009562b4 vwebp: Fix bug when Dispose then NoBlend frames
+423f2579 Fix up CMake to create targets.
+907208f9 Wait for all threads to be done in DecodeRemaining.
+4649b3c4 vwebp: Add background color display option
+78ad57a3 Fix bad glClearColor parameters
+da96d8d9 Allow for a non-initialized alpha decompressor in DoRemap.
+2563db47 fix rescaling rounding inaccuracy
+211f37ee fix endian problems in pattern copy
+5f0f5c07 Make sure partition #0 is read before VP8 data in IDecode.
+de98732b fix GetColorf() bug
+4338cd36 misc fixes in libwebpmux
+e00af13e fix signatures after a9ceda7ff1
+a9ceda7f Speed-up chunk list operations.
+2281bbf6 Merge "Better handling of bogus Huffman codes."
+39cb9aad Better handling of bogus Huffman codes.
+89cc9d37 Merge "fix read-overflow while parsing VP8X chunk"
+95fd6507 fix read-overflow while parsing VP8X chunk
+9e729fe1 Fix VP8IoTeardownHook being called twice on worker sync failure
+29fb8562 Merge "muxread,anmf: fail on multiple image chunks"
+eb82ce76 muxread,anmf: fail on multiple image chunks
+1344a2e9 fix alpha-filtering crash when image width is larger than radius
+be738c6d muxread,ChunkVerifyAndAssign: validate chunk_size
+2c70ad76 muxread,CreateInternal: fix riff size checks
+569001f1 Fix for thread race heap-use-after-free
+c56a02d9 Android.mk: use LOCAL_EXPORT_C_INCLUDES w/public libs
+15795596 CMakeLists.txt,cosmetics: normalize if() formatting
+1a44c233 Merge "cmake: add support for webpmux"
+e9569ad7 Merge "configure,*am,cosmetics: s/WANT_/BUILD_/"
+35c7de6f cmake: add support for webpmux
+0f25e61c WebpToSDL(): fix the return value in case of error
+5d8985de configure,*am,cosmetics: s/WANT_/BUILD_/
+895fd28f Merge "man/Makefile.am: add img2webp.1"
+5cf3e2af man/Makefile.am: add img2webp.1
+2a9de5b9 Add build rules for anim_diff & anim_dump utils.
+71ed73cf fix invalid check for buffer size
+af0e4fbb gif2webp: fix transcode of loop count=65535
+dce5d764 Limit memory allocation when reading invalid Huffman codes.
+f9df0081 Merge "cmake: quiet glut deprecation warnings on OS X"
+dc39b16f webpmux.1: correct grammar
+c7aa1264 cwebp.c: fix a missing \n
+53aa51e9 Merge tag 'v1.0.0'
+698b8844 update ChangeLog (tag: v1.0.0)
+8d510751 webp-container-spec: correct frame duration=0 note
+e6b2164e vwebp: Copy Chrome's behavior w/frame duration == 0
+094b3b28 cmake: quiet glut deprecation warnings on OS X
+71c39a06 webp-container-spec: correct frame duration=0 note
+fd3d5756 vwebp: Copy Chrome's behavior w/frame duration == 0
+b0c966fb Build vwebp from CMake.
+d20b7707 update ChangeLog (tag: v1.0.0-rc3)
+0d5fad46 add WEBP_DSP_INIT / WEBP_DSP_INIT_FUNC
+d77bf512 add WEBP_DSP_INIT / WEBP_DSP_INIT_FUNC
+c1cb86af fix 16b overflow in SSE2
+e577feb7 makefile.unix: add DEBUG flag for compiling w/ debug-symbol
+99be34b3 cwebp,get_disto: fix bpp output
+e122e511 cwebp,get_disto: fix bpp output
+f5565ca8 cmake: Make sure we use near-lossless by default.
+d898dc14 fix bug in WebPImport565: alpha value was not set
+1c8f358d Fix CMake with WASM.
+a0215fb7 webp_js: fix webp_js demo html
+882784b0 update ChangeLog (tag: v1.0.0-rc2)
+2f930e08 Revert "Use proper targets for CMake."
+8165e8fb Use proper targets for CMake.
+3f157dd5 Remove some very hard TODOs.
+abb47760 Merge "Use proper targets for CMake."
+cd758a17 {de,}mux/Makefile.am: add missing headers
+e155dda0 Use proper targets for CMake.
+b892b8ba makefile.unix,dist: use ascii for text output
+64a57d05 add -version option to anim_dump,anim_diff and img2webp
+994be82d Merge "Remove some very hard TODOs."
+4033e1d7 Remove some very hard TODOs.
+fc1b8e3a webp_js: fix webp_js demo html
+15aa48d9 update ChangeLog (tag: v1.0.0-rc1)
+e607dabc update AUTHORS
+38410c08 [CFI] Remove function pointer casts
+978eec25 [CFI] Remove function pointer casts
+c57b2736 bump version to 1.0.0
+cba28853 update NEWS
+c909d531 Merge "remove some deprecation warning on MacOSX"
+217443c7 remove some deprecation warning on MacOSX
+b672bdfa configure: quiet glut deprecation warnings on OS X
+daa9fcaf configure: use sdl-config if available
+dd174cae Merge "imagedec: support metadata reading for WebP image decoding"
+641cedcc imagedec: support metadata reading for WebP image decoding
+065b2ce1 anim_diff: add a couple missing newlines in Help()
+c4cc1147 Merge "gif2webp: force low duration frames to 100ms"
+09333097 gif2webp: force low duration frames to 100ms
+e03f0ec3 sharp_yuv: use 14b fixed-point precision for gamma
+b2db361c image_enc,WebPWritePNG: move locals after setjmp
+74e82ec6 Merge "WebPPictureDistortion: fix big-endian results order"
+645d04ca Merge "cwebp,get_disto: report bpp"
+120f58c3 Merge "lossless*sse2: improve non-const 16-bit vector creation"
+a7fe9412 WebPPictureDistortion: fix big-endian results order
+e26fe066 cwebp,get_disto: report bpp
+9df64e28 Merge changes Id5b4a1a4,Ia20ce844
+8043504f lossless*sse2: improve non-const 16-bit vector creation
+1e3dfc48 Import: extract condition from loop
+3b07d327 Import,RGBA: fix for BigEndian import
+551948e4 Remove unused argument in VP8LBitsEntropy.
+3005237a ReadWebP: fix for big-endian
+499c395a Merge "anim_diff: expose the -max_diff option"
+f69dcd69 Merge "remove WEBP_EXPERIMENTAL_FEATURES"
+07d884d5 anim_diff: expose the -max_diff option
+f4dd9256 remove WEBP_EXPERIMENTAL_FEATURES
+94a8377b extract the command-line parsing helpers to example_util
+fc09e6e2 PNM decoder: prevent unsupported depth=2 PAM case.
+6de58603 MIPS64: Fix defined-but-not-used errors with WEBP_REDUCE_CSP
+cbde5728 gif2webp: add support for reading from stdin
+cf1c5054 Add an SSE4 version of some lossless color transforms.
+45a8b5eb Fix lint error with man page.
+cff38e8f Merge "PNG decoder: handle gAMA chunk"
+59cb1a48 Merge "enable dc error-diffusion always"
+78318b30 PNG decoder: handle gAMA chunk
+664c21dd Merge "remove some TODOs"
+815652de enable dc error-diffusion always
+aec45cec remove some TODOs
+5715dfce fix block-count[] increment in case of large image
+c2d04f3e enable DC error-diffusion always for multi-pass
+96bf07c5 use DC error diffusion for U/V at low-quality
+1c59020b fix missing sse41 targets in makefile.unix
+7a8e814b cosmetics: s/color_space/colorspace/
+05f6fe24 upsampling: rm asserts w/REDUCE_CSP+OMIT_C_CODE
+b4cf5597 Merge "Upsampling SSE2/SSE4 speedup."
+ccbeb32c Makefile.vc: add missing sse41 files
+55403a9a Upsampling SSE2/SSE4 speedup.
+807b53c4 Implement the upsampling/yuv functions in SSE41
+84101a81 Fix wasm WebP compilation
+8bebd2a3 fix warning on MSVC
+a7f93fe3 webpmux: allow reading argument from a file
+b69f18a7 gif2webp.1: fix -loop_compatibility layout
+72d530c0 Merge "fix lossless decoding w/WEBP_REDUCE_SIZE"
+296c7dc4 fix lossless decoding w/WEBP_REDUCE_SIZE
+0d5d029c Merge "ImgIoUtilReadFile: fix file leak upon error"
+ae568ce7 ImgIoUtilReadFile: fix file leak upon error
+796b5a8a Merge tag 'v0.6.1'
+6b7a95fd update ChangeLog (tag: v0.6.1)
 f66955de WEBP_REDUCE_CSP: restrict colorspace support
+1af0df76 Merge "WEBP_REDUCE_CSP: restrict colorspace support"
+6de20df0 WEBP_REDUCE_CSP: restrict colorspace support
 a289d8e7 update ChangeLog (tag: v0.6.1-rc2)
 c10a493c vwebp: disable double buffering on windows & mac
 0d4466c2 webp_to_sdl.c: fix file mode
 1b27bf8b WEBP_REDUCE_SIZE: disable all rescaler code
 126be109 webpinfo: add -version option
+0df22b9e WEBP_REDUCE_SIZE: disable all rescaler code
 9add62b5 bump version to 0.6.1
 d3e26144 update NEWS
 2edda639 README: add webpinfo section

Makefile.am

@@ -2,7 +2,7 @@ ACLOCAL_AMFLAGS = -I m4
 SUBDIRS = src imageio man
 EXTRA_DIST = COPYING autogen.sh
 
-if WANT_EXTRAS
+if BUILD_EXTRAS
   SUBDIRS += extras
 endif
 

Makefile.vc

@@ -53,11 +53,6 @@ OUTDIR = ..\obj\
 OUTDIR = $(OBJDIR)
 !ENDIF
 
-!IF "$(HAVE_AVX2)" == "1"
-CFLAGS = $(CFLAGS) /DWEBP_HAVE_AVX2
-AVX2_FLAGS = /arch:AVX2
-!ENDIF
-
 ##############################################################
 # Runtime library configuration
 !IF "$(RTLIBCFG)" == "static"
@@ -227,11 +222,13 @@ DSP_DEC_OBJS = \
     $(DIROBJ)\dsp\upsampling_msa.obj \
     $(DIROBJ)\dsp\upsampling_neon.obj \
     $(DIROBJ)\dsp\upsampling_sse2.obj \
+    $(DIROBJ)\dsp\upsampling_sse41.obj \
     $(DIROBJ)\dsp\yuv.obj \
     $(DIROBJ)\dsp\yuv_mips32.obj \
     $(DIROBJ)\dsp\yuv_mips_dsp_r2.obj \
     $(DIROBJ)\dsp\yuv_neon.obj \
     $(DIROBJ)\dsp\yuv_sse2.obj \
+    $(DIROBJ)\dsp\yuv_sse41.obj \
 
 DSP_ENC_OBJS = \
     $(DIROBJ)\dsp\cost.obj \
@@ -239,7 +236,6 @@ DSP_ENC_OBJS = \
     $(DIROBJ)\dsp\cost_mips_dsp_r2.obj \
     $(DIROBJ)\dsp\cost_sse2.obj \
     $(DIROBJ)\dsp\enc.obj \
-    $(DIROBJ)\dsp\enc_avx2.obj \
     $(DIROBJ)\dsp\enc_mips32.obj \
     $(DIROBJ)\dsp\enc_mips_dsp_r2.obj \
     $(DIROBJ)\dsp\enc_msa.obj \
@@ -285,7 +281,6 @@ ENC_OBJS = \
     $(DIROBJ)\enc\backward_references_enc.obj \
     $(DIROBJ)\enc\config_enc.obj \
     $(DIROBJ)\enc\cost_enc.obj \
-    $(DIROBJ)\enc\delta_palettization_enc.obj \
     $(DIROBJ)\enc\filter_enc.obj \
     $(DIROBJ)\enc\frame_enc.obj \
     $(DIROBJ)\enc\histogram_enc.obj \
@@ -367,9 +362,11 @@ $(DIRBIN)\anim_dump.exe: $(EX_GIF_DEC_OBJS) $(LIBWEBPDEMUX) $(LIBWEBP)
 $(DIRBIN)\anim_dump.exe: $(IMAGEIO_ENC_OBJS)
 $(DIRBIN)\cwebp.exe: $(DIROBJ)\examples\cwebp.obj $(IMAGEIO_DEC_OBJS)
 $(DIRBIN)\cwebp.exe: $(IMAGEIO_UTIL_OBJS)
+$(DIRBIN)\cwebp.exe: $(LIBWEBPDEMUX)
 $(DIRBIN)\dwebp.exe: $(DIROBJ)\examples\dwebp.obj $(IMAGEIO_DEC_OBJS)
 $(DIRBIN)\dwebp.exe: $(IMAGEIO_ENC_OBJS)
 $(DIRBIN)\dwebp.exe: $(IMAGEIO_UTIL_OBJS)
+$(DIRBIN)\dwebp.exe: $(LIBWEBPDEMUX)
 $(DIRBIN)\gif2webp.exe: $(DIROBJ)\examples\gif2webp.obj $(EX_GIF_DEC_OBJS)
 $(DIRBIN)\gif2webp.exe: $(EX_UTIL_OBJS) $(IMAGEIO_UTIL_OBJS) $(LIBWEBPMUX)
 $(DIRBIN)\gif2webp.exe: $(LIBWEBP)
@@ -382,26 +379,24 @@ $(DIRBIN)\webpmux.exe: $(DIROBJ)\examples\webpmux.obj $(LIBWEBPMUX)
 $(DIRBIN)\webpmux.exe: $(EX_UTIL_OBJS) $(IMAGEIO_UTIL_OBJS) $(LIBWEBP)
 $(DIRBIN)\img2webp.exe: $(DIROBJ)\examples\img2webp.obj $(LIBWEBPMUX)
 $(DIRBIN)\img2webp.exe: $(IMAGEIO_DEC_OBJS)
-$(DIRBIN)\img2webp.exe: $(EX_UTIL_OBJS) $(IMAGEIO_UTIL_OBJS) $(LIBWEBP)
+$(DIRBIN)\img2webp.exe: $(EX_UTIL_OBJS) $(IMAGEIO_UTIL_OBJS)
+$(DIRBIN)\img2webp.exe: $(LIBWEBPDEMUX) $(LIBWEBP)
 $(DIRBIN)\get_disto.exe: $(DIROBJ)\extras\get_disto.obj
-$(DIRBIN)\get_disto.exe: $(IMAGEIO_DEC_OBJS) $(IMAGEIO_UTIL_OBJS) $(LIBWEBP)
+$(DIRBIN)\get_disto.exe: $(IMAGEIO_DEC_OBJS) $(IMAGEIO_UTIL_OBJS)
+$(DIRBIN)\get_disto.exe: $(LIBWEBPDEMUX) $(LIBWEBP)
 $(DIRBIN)\webp_quality.exe: $(DIROBJ)\extras\webp_quality.obj
 $(DIRBIN)\webp_quality.exe: $(IMAGEIO_UTIL_OBJS)
 $(DIRBIN)\webp_quality.exe: $(EXTRAS_OBJS) $(LIBWEBP)
 $(DIRBIN)\webpinfo.exe: $(DIROBJ)\examples\webpinfo.obj
 $(DIRBIN)\webpinfo.exe: $(IMAGEIO_DEC_OBJS)
-$(DIRBIN)\webpinfo.exe: $(EX_UTIL_OBJS) $(IMAGEIO_UTIL_OBJS) $(LIBWEBP)
+$(DIRBIN)\webpinfo.exe: $(EX_UTIL_OBJS) $(IMAGEIO_UTIL_OBJS)
+$(DIRBIN)\webpinfo.exe: $(LIBWEBPDEMUX) $(LIBWEBP)
 
 $(OUT_EXAMPLES): $(EX_UTIL_OBJS) $(LIBWEBP)
 $(EX_UTIL_OBJS) $(IMAGEIO_UTIL_OBJS): $(OUTPUT_DIRS)
 $(IMAGEIO_DEC_OBJS) $(IMAGEIO_ENC_OBJS) $(EXTRAS_OBJS): $(OUTPUT_DIRS)
 !ENDIF  # ARCH == ARM
 
-experimental:
-	$(MAKE) /f Makefile.vc \
-	    CFG=$(CFG) \
-	    CFLAGS="$(CFLAGS) /DWEBP_EXPERIMENTAL_FEATURES" /$(MAKEFLAGS)
-
 $(LIBWEBPDECODER): $(LIBWEBPDECODER_OBJS)
 $(LIBWEBP): $(LIBWEBP_OBJS)
 $(LIBWEBPMUX): $(LIBWEBPMUX_OBJS)
@@ -453,9 +448,6 @@ $(DIROBJ)\$(DLLINC):
 .SUFFIXES: .c .obj .res .exe
 # File-specific flag builds. Note batch rules take precedence over wildcards,
 # so for now name each file individually.
-$(DIROBJ)\dsp\enc_avx2.obj: src\dsp\enc_avx2.c
-	$(CC) $(CFLAGS) $(AVX2_FLAGS) /Fd$(LIBWEBP_PDBNAME) /Fo$(DIROBJ)\dsp\ \
-	  src\dsp\$(@B).c
 $(DIROBJ)\examples\anim_diff.obj: examples\anim_diff.c
 	$(CC) $(CFLAGS) /DWEBP_HAVE_GIF /Fd$(LIBWEBP_PDBNAME) \
 	  /Fo$(DIROBJ)\examples\ examples\$(@B).c

NEWS

@@ -1,3 +1,25 @@
+- 11/2/2018: version 1.0.1
+  This is a binary compatible release.
+  * lossless encoder speedups
+  * big-endian fix for alpha decoding (issue #393)
+  * gif2webp fix for loop count=65535 transcode (issue #382)
+  * further security related hardening in libwebp & libwebpmux
+    (issues #383, #385, #386, #387, #388, #391)
+    (oss-fuzz #9099, #9100, #9105, #9106, #9111, #9112, #9119, #9123, #9170,
+              #9178, #9179, #9183, #9186, #9191, #9364, #9417, #9496, #10349,
+              #10423, #10634, #10700, #10838, #10922, #11021, #11088, #11152)
+  * miscellaneous bug & build fixes (issues #381, #394, #396, #397, #400)
+
+- 4/2/2018: version 1.0.0
+  This is a binary compatible release.
+  * lossy encoder improvements to avoid chroma shifts in various circumstances
+    (issues #308, #340)
+  * big-endian fixes for decode, RGBA import and WebPPictureDistortion
+  Tool updates:
+    gifwebp, anim_diff - default duration behavior (<= 10ms) changed to match
+                         web browsers, transcoding tools (issue #379)
+    img2webp, webpmux - allow options to be passed in via a file (issue #355)
+
 - 11/24/2017: version 0.6.1
   This is a binary compatible release.
   * lossless performance and compression improvements + a new 'cruncher' mode

README

@@ -4,7 +4,7 @@
           \__\__/\____/\_____/__/ ____  ___
                 / _/ /    \    \ /  _ \/ _/
                /  \_/   / /   \ \   __/  \__
-               \____/____/\_____/_____/____/v0.6.1
+               \____/____/\_____/_____/____/v1.0.1
 
 Description:
 ============
@@ -402,12 +402,14 @@ Options are:
   -nofilter .... disable in-loop filtering
   -dither <int>  dithering strength (0..100), default=50
   -noalphadither disable alpha plane dithering
+  -usebgcolor .. display background color
   -mt .......... use multi-threading
   -info ........ print info
   -h ........... this help message
 
 Keyboard shortcuts:
   'c' ................ toggle use of color profile
+  'b' ................ toggle background color display
   'i' ................ overlay file information
   'd' ................ disable blending & disposal (debug)
   'q' / 'Q' / ESC .... quit
@@ -458,6 +460,7 @@ File-level options (only used at the start of compression):
  -mixed ............... use mixed lossy/lossless automatic mode
  -v ................... verbose mode
  -h ................... this help
+ -version ............. print version number and exit
 
 Per-frame options (only used for subsequent images input):
  -d <int> ............. frame duration in ms (default: 100)
@@ -469,6 +472,9 @@ Per-frame options (only used for subsequent images input):
 example: img2webp -loop 2 in0.png -lossy in1.jpg
                   -d 80 in2.tiff -o out.webp
 
+Note: if a single file name is passed as the argument, the arguments will be
+tokenized from this file. The file name must not start with the character '-'.
+
 Animated GIF conversion:
 ========================
 Animated GIF files can be converted to WebP files with animation using the
@@ -524,6 +530,11 @@ Options:
   -min_psnr <float> ... minimum per-frame PSNR
   -raw_comparison ..... if this flag is not used, RGB is
                         premultiplied before comparison
+  -max_diff <int> ..... maximum allowed difference per channel
+                        between corresponding pixels in subsequent
+                        frames
+  -h .................. this help
+  -version ............ print version number and exit
 
 Building:
 ---------

README.mux

@@ -1,7 +1,7 @@
           __   __  ____  ____  ____  __ __  _     __ __
          /  \\/  \/  _ \/  _ \/  _ \/  \  \/ \___/_ / _\
          \       /   __/  _  \   __/      /  /  (_/  /__
-          \__\__/\_____/_____/__/  \__//_/\_____/__/___/v0.4.1
+          \__\__/\_____/_____/__/  \__//_/\_____/__/___/v1.0.1
 
 
 Description:
@@ -33,6 +33,7 @@ Usage: webpmux -get GET_OPTIONS INPUT -o OUTPUT
        webpmux -info INPUT
        webpmux [-h|-help]
        webpmux -version
+       webpmux argument_file_name
 
 GET_OPTIONS:
  Extract relevant data:
@@ -92,6 +93,9 @@ INPUT & OUTPUT are in WebP format.
 Note: The nature of EXIF, XMP and ICC data is not checked and is assumed to be
 valid.
 
+Note: if a single file name is passed as the argument, the arguments will be
+tokenized from this file. The file name must not start with the character '-'.
+
 Visualization tool:
 ===================
 
@@ -207,6 +211,35 @@ Code example:
 For a detailed AnimEncoder API reference, please refer to the header file
 (src/webp/mux.h).
 
+AnimDecoder API:
+================
+This AnimDecoder API allows decoding (possibly) animated WebP images.
+
+Code Example:
+
+  WebPAnimDecoderOptions dec_options;
+  WebPAnimDecoderOptionsInit(&dec_options);
+  // Tune 'dec_options' as needed.
+  WebPAnimDecoder* dec = WebPAnimDecoderNew(webp_data, &dec_options);
+  WebPAnimInfo anim_info;
+  WebPAnimDecoderGetInfo(dec, &anim_info);
+  for (uint32_t i = 0; i < anim_info.loop_count; ++i) {
+    while (WebPAnimDecoderHasMoreFrames(dec)) {
+      uint8_t* buf;
+      int timestamp;
+      WebPAnimDecoderGetNext(dec, &buf, &timestamp);
+      // ... (Render 'buf' based on 'timestamp').
+      // ... (Do NOT free 'buf', as it is owned by 'dec').
+    }
+    WebPAnimDecoderReset(dec);
+  }
+  const WebPDemuxer* demuxer = WebPAnimDecoderGetDemuxer(dec);
+  // ... (Do something using 'demuxer'; e.g. get EXIF/XMP/ICC data).
+  WebPAnimDecoderDelete(dec);
+
+For a detailed AnimDecoder API reference, please refer to the header file
+(src/webp/demux.h).
+
 
 Bugs:
 =====

README.webp_js

@@ -32,7 +32,8 @@ using Emscripten and CMake.
    webp.js.mem files generated.
 
 The callable JavaScript function is WebPToSDL(), which decodes a raw WebP
-bitstream into a canvas. See webp_js/index.html for a simple usage sample.
+bitstream into a canvas. See webp_js/index.html for a simple usage sample
+(see below for instructions).
 
 Demo HTML page:
 ===============

build.gradle

@@ -152,11 +152,13 @@ model {
             include "upsampling_msa.c"
             include "upsampling_neon.$NEON"
             include "upsampling_sse2.c"
+            include "upsampling_sse41.c"
             include "yuv.c"
             include "yuv_mips32.c"
             include "yuv_mips_dsp_r2.c"
             include "yuv_neon.$NEON"
             include "yuv_sse2.c"
+            include "yuv_sse41.c"
             srcDir "src/utils"
             include "bit_reader_utils.c"
             include "color_cache_utils.c"
@@ -173,7 +175,6 @@ model {
             include "cost_mips_dsp_r2.c"
             include "cost_sse2.c"
             include "enc.c"
-            include "enc_avx2.c"
             include "enc_mips32.c"
             include "enc_mips_dsp_r2.c"
             include "enc_msa.c"
@@ -196,7 +197,6 @@ model {
             include "backward_references_enc.c"
             include "config_enc.c"
             include "cost_enc.c"
-            include "delta_palettization_enc.c"
             include "filter_enc.c"
             include "frame_enc.c"
             include "histogram_enc.c"
@@ -288,6 +288,7 @@ model {
     imagedec(NativeLibrarySpec) {
       binaries {
         all {
+          lib library: "webpdemux", linkage: "static"
           lib library: "webp", linkage: "static"
         }
       }
@@ -330,6 +331,7 @@ model {
           lib library: "example_util", linkage: "static"
           lib library: "imagedec", linkage: "static"
           lib library: "imageio_util", linkage: "static"
+          lib library: "webpdemux", linkage: "static"
           lib library: "webp", linkage: "static"
         }
       }
@@ -350,6 +352,7 @@ model {
           lib library: "imagedec", linkage: "static"
           lib library: "imageenc", linkage: "static"
           lib library: "imageio_util", linkage: "static"
+          lib library: "webpdemux", linkage: "static"
           lib library: "webp"
         }
       }
@@ -389,6 +392,7 @@ model {
           lib library: "imagedec", linkage: "static"
           lib library: "imageio_util", linkage: "static"
           lib library: "webpmux", linkage: "static"
+          lib library: "webpdemux", linkage: "static"
           lib library: "webp"
         }
       }

cmake/WebPConfig.cmake.in

@@ -3,17 +3,9 @@ set(WEBP_VERSION ${WebP_VERSION})
 
 @PACKAGE_INIT@
 
-if(@WEBP_USE_THREAD@)
-  include(CMakeFindDependencyMacro)
-  find_dependency(Threads REQUIRED)
-endif()
+include ("${CMAKE_CURRENT_LIST_DIR}/@PROJECT_NAME@Targets.cmake")
 
-include("${CMAKE_CURRENT_LIST_DIR}/WebPTargets.cmake")
-
-set_and_check(WebP_INCLUDE_DIR "@PACKAGE_CMAKE_INSTALL_INCLUDEDIR@")
-set(WebP_INCLUDE_DIRS ${WebP_INCLUDE_DIR})
-set(WEBP_INCLUDE_DIRS ${WebP_INCLUDE_DIR})
+set(WebP_INCLUDE_DIRS "webp")
+set(WEBP_INCLUDE_DIRS ${WebP_INCLUDE_DIRS})
 set(WebP_LIBRARIES "@INSTALLED_LIBRARIES@")
 set(WEBP_LIBRARIES "${WebP_LIBRARIES}")
-
-check_required_components(WebP)

cmake/config.h.in

@@ -103,12 +103,6 @@
 /* Version number of package */
 #cmakedefine VERSION "@VERSION@"
 
-/* Enable experimental code */
-#cmakedefine WEBP_EXPERIMENTAL_FEATURES 1
-
-/* Set to 1 if AVX2 is supported */
-#cmakedefine WEBP_HAVE_AVX2 1
-
 /* Set to 1 if GIF library is installed */
 #cmakedefine WEBP_HAVE_GIF 1
 

cmake/cpu.cmake

@@ -1,4 +1,4 @@
-## Check for SIMD extensions.
+# Check for SIMD extensions.
 include(CMakePushCheckState)
 
 function(webp_check_compiler_flag WEBP_SIMD_FLAG ENABLE_SIMD)
@@ -18,8 +18,7 @@ function(webp_check_compiler_flag WEBP_SIMD_FLAG ENABLE_SIMD)
         #endif
         return 0;
       }
-    " WEBP_HAVE_FLAG_${WEBP_SIMD_FLAG}
-  )
+    " WEBP_HAVE_FLAG_${WEBP_SIMD_FLAG})
   cmake_pop_check_state()
   if(WEBP_HAVE_FLAG_${WEBP_SIMD_FLAG})
     set(WEBP_HAVE_${WEBP_SIMD_FLAG} 1 PARENT_SCOPE)
@@ -29,16 +28,18 @@ function(webp_check_compiler_flag WEBP_SIMD_FLAG ENABLE_SIMD)
 endfunction()
 
 # those are included in the names of WEBP_USE_* in c++ code.
-set(WEBP_SIMD_FLAGS "SSE2;SSE41;AVX2;MIPS32;MIPS_DSP_R2;NEON;MSA")
-set(WEBP_SIMD_FILE_EXTENSIONS "_sse2.c;_sse41.c;_avx2.c;_mips32.c;_mips_dsp_r2.c;_neon.c;_msa.c")
+set(WEBP_SIMD_FLAGS "SSE41;SSE2;MIPS32;MIPS_DSP_R2;NEON;MSA")
+set(WEBP_SIMD_FILE_EXTENSIONS
+    "_sse41.c;_sse2.c;_mips32.c;_mips_dsp_r2.c;_neon.c;_msa.c")
 if(MSVC)
-  # MSVC does not have a SSE4 flag but AVX2 support implies
-  # SSE4 support.
-  set(SIMD_ENABLE_FLAGS "/arch:SSE2;/arch:AVX2;/arch:AVX2;;;;")
+  # MSVC does not have a SSE4 flag but AVX support implies SSE4 support.
+  set(SIMD_ENABLE_FLAGS "/arch:AVX;/arch:SSE2;;;;")
   set(SIMD_DISABLE_FLAGS)
 else()
-  set(SIMD_ENABLE_FLAGS "-msse2;-msse4.1;-mavx2;-mips32;-mdspr2;-mfpu=neon;-mmsa")
-  set(SIMD_DISABLE_FLAGS "-mno-sse2;-mno-sse4.1;-mno-avx2;;-mno-dspr2;;-mno-msa")
+  set(SIMD_ENABLE_FLAGS
+      "-msse4.1;-msse2;-mips32;-mdspr2;-mfpu=neon;-mmsa")
+  set(SIMD_DISABLE_FLAGS
+      "-mno-sse4.1;-mno-sse2;;-mno-dspr2;;-mno-msa")
 endif()
 
 set(WEBP_SIMD_FILES_TO_NOT_INCLUDE)
@@ -47,16 +48,16 @@ set(WEBP_SIMD_FLAGS_TO_INCLUDE)
 
 if(${ANDROID})
   if(${ANDROID_ABI} STREQUAL "armeabi-v7a")
-    # This is because Android studio uses the configuration
-    # "-march=armv7-a -mfloat-abi=softfp -mfpu=vfpv3-d16"
-    # that does not trigger neon optimizations but should
-    # (as this configuration does not exist anymore).
+    # This is because Android studio uses the configuration "-march=armv7-a
+    # -mfloat-abi=softfp -mfpu=vfpv3-d16" that does not trigger neon
+    # optimizations but should (as this configuration does not exist anymore).
     set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mfpu=neon ")
   endif()
 endif()
 
 list(LENGTH WEBP_SIMD_FLAGS WEBP_SIMD_FLAGS_LENGTH)
 math(EXPR WEBP_SIMD_FLAGS_RANGE "${WEBP_SIMD_FLAGS_LENGTH} - 1")
+unset(HIGHEST_SSE_FLAG)
 
 foreach(I_SIMD RANGE ${WEBP_SIMD_FLAGS_RANGE})
   list(GET WEBP_SIMD_FLAGS ${I_SIMD} WEBP_SIMD_FLAG)
@@ -72,18 +73,28 @@ foreach(I_SIMD RANGE ${WEBP_SIMD_FLAGS_RANGE})
     set(CMAKE_REQUIRED_FLAGS ${SIMD_COMPILE_FLAG})
     webp_check_compiler_flag(${WEBP_SIMD_FLAG} ${WEBP_ENABLE_SIMD})
   else()
-    set(SIMD_COMPILE_FLAG " ")
+    if(MSVC)
+      list(GET SIMD_ENABLE_FLAGS ${I_SIMD} SIMD_COMPILE_FLAG)
+    else()
+      set(SIMD_COMPILE_FLAG " ")
+    endif()
   endif()
   # Check which files we should include or not.
   list(GET WEBP_SIMD_FILE_EXTENSIONS ${I_SIMD} WEBP_SIMD_FILE_EXTENSION)
   file(GLOB SIMD_FILES "${CMAKE_CURRENT_LIST_DIR}/../"
-    "src/dsp/*${WEBP_SIMD_FILE_EXTENSION}"
-  )
+            "src/dsp/*${WEBP_SIMD_FILE_EXTENSION}")
   if(WEBP_HAVE_${WEBP_SIMD_FLAG})
+    if(${I_SIMD} LESS 2 AND NOT HIGHEST_SSE_FLAG)
+      set(HIGHEST_SSE_FLAG ${SIMD_COMPILE_FLAG})
+    endif()
     # Memorize the file and flags.
     foreach(FILE ${SIMD_FILES})
       list(APPEND WEBP_SIMD_FILES_TO_INCLUDE ${FILE})
-      list(APPEND WEBP_SIMD_FLAGS_TO_INCLUDE ${SIMD_COMPILE_FLAG})
+      if(${I_SIMD} LESS 2)
+        list(APPEND WEBP_SIMD_FLAGS_TO_INCLUDE ${HIGHEST_SSE_FLAG})
+      else()
+        list(APPEND WEBP_SIMD_FLAGS_TO_INCLUDE ${SIMD_COMPILE_FLAG})
+      endif()
     endforeach()
   else()
     # Remove the file from the list.
@@ -107,10 +118,10 @@ foreach(I_SIMD RANGE ${WEBP_SIMD_FLAGS_RANGE})
           endif()
           set(CMAKE_REQUIRED_DEFINITIONS ${SIMD_COMPILE_FLAG})
           check_c_source_compiles("int main(void) {return 0;}"
-            FLAG_${SIMD_COMPILE_FLAG}
-            FAIL_REGEX "warning: argument unused during compilation:"
-            ${COMMON_PATTERNS}
-          )
+                                  FLAG_${SIMD_COMPILE_FLAG}
+                                  FAIL_REGEX
+                                  "warning: argument unused during compilation:"
+                                  ${COMMON_PATTERNS})
           if(NOT FLAG_${SIMD_COMPILE_FLAG})
             unset(HAS_COMPILE_FLAG CACHE)
           endif()

cmake/deps.cmake

@@ -1,33 +1,27 @@
 # Generate the config.h to compile with specific intrinsics / libs.
 
-## Check for compiler options.
+# Check for compiler options.
 include(CheckCSourceCompiles)
 check_c_source_compiles("
     int main(void) {
       (void)__builtin_bswap16(0);
       return 0;
     }
-  "
-  HAVE_BUILTIN_BSWAP16
-)
+  " HAVE_BUILTIN_BSWAP16)
 check_c_source_compiles("
     int main(void) {
       (void)__builtin_bswap32(0);
       return 0;
     }
-  "
-  HAVE_BUILTIN_BSWAP32
-)
+  " HAVE_BUILTIN_BSWAP32)
 check_c_source_compiles("
     int main(void) {
       (void)__builtin_bswap64(0);
       return 0;
     }
-  "
-  HAVE_BUILTIN_BSWAP64
-)
+  " HAVE_BUILTIN_BSWAP64)
 
-## Check for libraries.
+# Check for libraries.
 find_package(Threads)
 if(Threads_FOUND)
   if(CMAKE_USE_PTHREADS_INIT)
@@ -40,8 +34,7 @@ if(Threads_FOUND)
           int attr = ${PTHREAD_TEST};
           return attr;
         }
-      " ${PTHREAD_TEST}
-    )
+      " ${PTHREAD_TEST})
   endforeach()
   list(APPEND WEBP_DEP_LIBRARIES ${CMAKE_THREAD_LIBS_INIT})
 endif()
@@ -51,15 +44,20 @@ set(WEBP_USE_THREAD ${Threads_FOUND})
 set(LT_OBJDIR ".libs/")
 
 # Only useful for vwebp, so useless for now.
-# find_package(OpenGL)
-# set(WEBP_HAVE_GL ${OPENGL_FOUND})
-# set(WEBP_DEP_INCLUDE_DIRS ${WEBP_DEP_INCLUDE_DIRS} ${OPENGL_INCLUDE_DIRS})
-# set(WEBP_DEP_LIBRARIES ${WEBP_DEP_LIBRARIES} ${OPENGL_LIBRARIES})
+find_package(OpenGL)
+set(WEBP_HAVE_GL ${OPENGL_FOUND})
 
-# Find the standard C math library.
-find_library(MATH_LIBRARY NAMES m)
-if(MATH_LIBRARY)
-  list(APPEND WEBP_DEP_LIBRARIES ${MATH_LIBRARY})
+# Check if we need to link to the C math library. We do not look for it as it is
+# not found when cross-compiling, while it is here.
+check_c_source_compiles("
+    #include <math.h>
+    int main(int argc, char** argv) {
+      return (int)pow(argc, 2.5);
+    }
+  " HAVE_MATH_LIBRARY)
+if(NOT HAVE_MATH_LIBRARY)
+  message(STATUS "Adding -lm flag.")
+  list(APPEND WEBP_DEP_LIBRARIES m)
 endif()
 
 # Find the standard image libraries.
@@ -70,8 +68,8 @@ foreach(I_LIB PNG JPEG TIFF)
   set(WEBP_HAVE_${I_LIB} ${${I_LIB}_FOUND})
   if(${I_LIB}_FOUND)
     list(APPEND WEBP_DEP_IMG_LIBRARIES ${${I_LIB}_LIBRARIES})
-    list(APPEND WEBP_DEP_IMG_INCLUDE_DIRS
-         ${${I_LIB}_INCLUDE_DIR} ${${I_LIB}_INCLUDE_DIRS})
+    list(APPEND WEBP_DEP_IMG_INCLUDE_DIRS ${${I_LIB}_INCLUDE_DIR}
+                ${${I_LIB}_INCLUDE_DIRS})
   endif()
 endforeach()
 if(WEBP_DEP_IMG_INCLUDE_DIRS)
@@ -87,8 +85,8 @@ set(WEBP_HAVE_GIF ${GIF_FOUND})
 if(GIF_FOUND)
   # GIF find_package only locates the header and library, it doesn't fail
   # compile tests when detecting the version, but falls back to 3 (as of at
-  # least cmake 3.7.2). Make sure the library links to avoid incorrect
-  # detection when cross compiling.
+  # least cmake 3.7.2). Make sure the library links to avoid incorrect detection
+  # when cross compiling.
   cmake_push_check_state()
   set(CMAKE_REQUIRED_LIBRARIES ${GIF_LIBRARIES})
   set(CMAKE_REQUIRED_INCLUDES ${GIF_INCLUDE_DIR})
@@ -98,8 +96,7 @@ if(GIF_FOUND)
         (void)DGifOpenFileHandle;
         return 0;
       }
-      " GIF_COMPILES
-  )
+      " GIF_COMPILES)
   cmake_pop_check_state()
   if(GIF_COMPILES)
     list(APPEND WEBP_DEP_GIF_LIBRARIES ${GIF_LIBRARIES})
@@ -109,7 +106,7 @@ if(GIF_FOUND)
   endif()
 endif()
 
-## Check for specific headers.
+# Check for specific headers.
 include(CheckIncludeFiles)
 check_include_files("stdlib.h;stdarg.h;string.h;float.h" STDC_HEADERS)
 check_include_files(dlfcn.h HAVE_DLFCN_H)
@@ -131,25 +128,32 @@ check_include_files(windows.h HAVE_WINDOWS_H)
 
 # Windows specifics
 if(HAVE_WINCODEC_H)
-  list(APPEND WEBP_DEP_LIBRARIES shlwapi ole32 windowscodecs)
+  list(APPEND WEBP_DEP_LIBRARIES
+              shlwapi
+              ole32
+              windowscodecs)
 endif()
 
-## Check for SIMD extensions.
+# Check for SIMD extensions.
 include(${CMAKE_CURRENT_LIST_DIR}/cpu.cmake)
 
-## Define extra info.
+# Define extra info.
 set(PACKAGE ${PROJECT_NAME})
 set(PACKAGE_NAME ${PROJECT_NAME})
 
 # Read from configure.ac.
 file(READ ${CMAKE_CURRENT_SOURCE_DIR}/configure.ac CONFIGURE_AC)
-string(REGEX MATCHALL "\\[([0-9a-z\\.:/]*)\\]"
-  CONFIGURE_AC_PACKAGE_INFO ${CONFIGURE_AC}
-)
+string(REGEX MATCHALL
+             "\\[([0-9a-z\\.:/]*)\\]"
+             CONFIGURE_AC_PACKAGE_INFO
+             ${CONFIGURE_AC})
 function(strip_bracket VAR)
   string(LENGTH ${${VAR}} TMP_LEN)
   math(EXPR TMP_LEN ${TMP_LEN}-2)
-  string(SUBSTRING ${${VAR}} 1 ${TMP_LEN} TMP_SUB)
+  string(SUBSTRING ${${VAR}}
+                   1
+                   ${TMP_LEN}
+                   TMP_SUB)
   set(${VAR} ${TMP_SUB} PARENT_SCOPE)
 endfunction()
 

configure.ac

@@ -1,4 +1,4 @@
-AC_INIT([libwebp], [0.6.1],
+AC_INIT([libwebp], [1.0.1],
         [https://bugs.chromium.org/p/webp],,
         [http://developers.google.com/speed/webp])
 AC_CANONICAL_HOST
@@ -122,31 +122,6 @@ AS_IF([test "$GCC" = "yes" ], [
 AC_SUBST([AM_CFLAGS])
 
 dnl === Check for machine specific flags
-AC_ARG_ENABLE([avx2],
-              AS_HELP_STRING([--disable-avx2],
-                             [Disable detection of AVX2 support
-                              @<:@default=auto@:>@]))
-
-AS_IF([test "x$enable_avx2" != "xno" -a "x$enable_sse4_1" != "xno" \
-         -a "x$enable_sse2" != "xno"], [
-  AVX2_CFLAGS="$INTRINSICS_CFLAGS $AVX2_FLAGS"
-  TEST_AND_ADD_CFLAGS([AVX2_FLAGS], [-mavx2])
-  AS_IF([test -n "$AVX2_FLAGS"], [
-    SAVED_CFLAGS=$CFLAGS
-    CFLAGS="$CFLAGS $AVX2_FLAGS"
-    AC_CHECK_HEADER([immintrin.h],
-                    [AC_DEFINE(WEBP_HAVE_AVX2, [1],
-                     [Set to 1 if AVX2 is supported])],
-                    [AVX2_FLAGS=""],
-                    dnl it's illegal to directly include avx2intrin.h, but it's
-                    dnl included conditionally in immintrin.h, tricky!
-                    [#ifndef __AVX2__
-                     #error avx2 is not enabled
-                     #endif
-                    ])
-    CFLAGS=$SAVED_CFLAGS])
-  AC_SUBST([AVX2_FLAGS])])
-
 AC_ARG_ENABLE([sse4.1],
               AS_HELP_STRING([--disable-sse4.1],
                              [Disable detection of SSE4.1 support
@@ -347,6 +322,8 @@ AS_IF([test "x$enable_gl" != "xno"], [
       # override with --with-gl*
       glut_cflags="$glut_cflags|-framework GLUT -framework OpenGL"
       glut_ldflags="$glut_ldflags|-framework GLUT -framework OpenGL"
+      # quiet deprecation warnings for glut
+      TEST_AND_ADD_CFLAGS([AM_CFLAGS], [-Wno-deprecated-declarations])
       ;;
   esac
 
@@ -443,31 +420,54 @@ AC_ARG_ENABLE([sdl],
                               @<:@default=auto@:>@]))
 AS_IF([test "x$enable_sdl" != "xno"], [
   CLEAR_LIBVARS([SDL])
+  AC_PATH_PROGS([LIBSDL_CONFIG], [sdl-config])
+  if test -n "$LIBSDL_CONFIG"; then
+    SDL_INCLUDES=`$LIBSDL_CONFIG --cflags`
+    SDL_LIBS="`$LIBSDL_CONFIG --libs`"
+  fi
+
   WITHLIB_OPTION([sdl], [SDL])
 
   sdl_header="no"
   LIBCHECK_PROLOGUE([SDL])
-  AC_CHECK_HEADER([SDL/SDL.h], [sdl_header="SDL_SDL.h"],
+  AC_CHECK_HEADER([SDL/SDL.h], [sdl_header="SDL/SDL.h"],
                   [AC_CHECK_HEADER([SDL.h], [sdl_header="SDL.h"],
                   [AC_MSG_WARN(SDL library not available - no sdl.h)])])
   if test x"$sdl_header" != "xno"; then
-    AC_CHECK_LIB(SDL, SDL_Init,
-                 [SDL_LIBS="-lSDL"
-                  SDL_INCLUDES="-DWEBP_HAVE_SDL"
-                  AC_DEFINE(WEBP_HAVE_SDL, [1],
-                            [Set to 1 if SDL library is installed])
-                  sdl_support=yes
-                 ],
-                 AC_MSG_WARN(Optional SDL library not found),
-                 [$MATH_LIBS])
+    AC_LANG_PUSH(C)
+    SDL_SAVED_LIBS="$LIBS"
+    for lib in "" "-lSDL" "-lSDLmain -lSDL"; do
+      LIBS="$SDL_SAVED_LIBS $lib"
+      # Perform a full link to ensure SDL_main is resolved if needed.
+      AC_LINK_IFELSE(
+        [AC_LANG_SOURCE([
+           #include <$sdl_header>
+           int main(int argc, char** argv) {
+             SDL_Init(0);
+             return 0;
+           }])],
+        [SDL_LIBS="$LDFLAGS $LIBS"
+         SDL_INCLUDES="$SDL_INCLUDES -DWEBP_HAVE_SDL"
+         AC_DEFINE(WEBP_HAVE_SDL, [1],
+                   [Set to 1 if SDL library is installed])
+         sdl_support=yes]
+      )
+      if test x"$sdl_support" = "xyes"; then
+        break
+      fi
+    done
+    # LIBS is restored by LIBCHECK_EPILOGUE
+    AC_LANG_POP
     if test x"$sdl_header" = "xSDL.h"; then
       SDL_INCLUDES="$SDL_INCLUDES -DWEBP_HAVE_JUST_SDL_H"
     fi
   fi
   LIBCHECK_EPILOGUE([SDL])
 
-  if test "$sdl_support" = "yes"; then
+  if test x"$sdl_support" = "xyes"; then
     build_vwebp_sdl=yes
+  else
+    AC_MSG_WARN(Optional SDL library not found)
   fi
 ])
 
@@ -601,7 +601,7 @@ AS_IF([test "x$enable_gif" != "xno"], [
 AM_CONDITIONAL([BUILD_ANIMDIFF], [test "${build_anim_diff}" = "yes"])
 AM_CONDITIONAL([BUILD_GIF2WEBP], [test "${build_gif2webp}" = "yes"])
 
-if test "$enable_libwebpmux" = "yes"; then
+if test "$enable_libwebpdemux" = "yes" -a "$enable_libwebpmux" = "yes"; then
   build_img2webp=yes
 fi
 AM_CONDITIONAL([BUILD_IMG2WEBP], [test "${build_img2webp}" = "yes"])
@@ -676,19 +676,6 @@ fi
 AC_MSG_RESULT(${enable_swap_16bit_csp-no})
 AC_SUBST(USE_SWAP_16BIT_CSP)
 
-dnl === If --enable-experimental is defined, add -DWEBP_EXPERIMENTAL_FEATURES
-
-USE_EXPERIMENTAL_CODE=""
-AC_MSG_CHECKING(if --enable-experimental option is specified)
-AC_ARG_ENABLE([experimental], AS_HELP_STRING([--enable-experimental],
-                                             [Activate experimental features]))
-if test "$enable_experimental" = "yes"; then
-  AC_DEFINE(WEBP_EXPERIMENTAL_FEATURES, [1], [Enable experimental code])
-  USE_EXPERIMENTAL_CODE="-DWEBP_EXPERIMENTAL_FEATURES"
-fi
-AC_MSG_RESULT(${enable_experimental-no})
-AC_SUBST(USE_EXPERIMENTAL_CODE)
-
 dnl === If --disable-near-lossless is defined, add -DWEBP_NEAR_LOSSLESS=0
 
 AC_DEFINE(WEBP_NEAR_LOSSLESS, [1], [Enable near lossless encoding])
@@ -710,15 +697,16 @@ AC_ARG_ENABLE([libwebpmux],
               AS_HELP_STRING([--enable-libwebpmux],
                              [Build libwebpmux @<:@default=no@:>@]))
 AC_MSG_RESULT(${enable_libwebpmux-no})
-AM_CONDITIONAL([WANT_MUX], [test "$enable_libwebpmux" = "yes"])
+AM_CONDITIONAL([BUILD_MUX], [test "$enable_libwebpmux" = "yes"])
 
 dnl === Check whether libwebpdemux should be built
 AC_MSG_CHECKING(whether libwebpdemux is to be built)
 AC_ARG_ENABLE([libwebpdemux],
-              AS_HELP_STRING([--enable-libwebpdemux],
-                             [Build libwebpdemux @<:@default=no@:>@]))
+              AS_HELP_STRING([--disable-libwebpdemux],
+                             [Disable libwebpdemux @<:@default=no@:>@]),
+              [], [enable_libwebpdemux=yes])
 AC_MSG_RESULT(${enable_libwebpdemux-no})
-AM_CONDITIONAL([WANT_DEMUX], [test "$enable_libwebpdemux" = "yes"])
+AM_CONDITIONAL([BUILD_DEMUX], [test "$enable_libwebpdemux" = "yes"])
 
 dnl === Check whether decoder library should be built.
 AC_MSG_CHECKING(whether decoder library is to be built)
@@ -734,7 +722,7 @@ AC_ARG_ENABLE([libwebpextras],
               AS_HELP_STRING([--enable-libwebpextras],
                              [Build libwebpextras @<:@default=no@:>@]))
 AC_MSG_RESULT(${enable_libwebpextras-no})
-AM_CONDITIONAL([WANT_EXTRAS], [test "$enable_libwebpextras" = "yes"])
+AM_CONDITIONAL([BUILD_EXTRAS], [test "$enable_libwebpextras" = "yes"])
 
 dnl =========================
 
@@ -765,14 +753,14 @@ libwebpmux: ${enable_libwebpmux-no}
 libwebpextras: ${enable_libwebpextras-no}
 
 Tools:
-cwebp : yes
+cwebp : ${enable_libwebpdemux-no}
   Input format support
   ====================
   JPEG : ${jpeg_support-no}
   PNG  : ${png_support-no}
   TIFF : ${tiff_support-no}
   WIC  : ${wic_support-no}
-dwebp : yes
+dwebp : ${enable_libwebpdemux-no}
   Output format support
   =====================
   PNG  : ${png_support-no}

doc/webp-container-spec.txt

@@ -446,8 +446,9 @@ Frame Height Minus One: 24 bits (_uint24_)
 Frame Duration: 24 bits (_uint24_)
 
 : The time to wait before displaying the next frame, in 1 millisecond units.
-  In particular, frame duration of 0 is useful when one wants to update
-  multiple areas of the canvas at once during the animation.
+  Note the interpretation of frame duration of 0 (and often <= 10) is
+  implementation defined. Many tools and browsers assign a minimum duration
+  similar to GIF.
 
 Reserved: 6 bits
 

examples/Android.mk

@@ -26,8 +26,7 @@ LOCAL_SRC_FILES := \
     cwebp.c \
 
 LOCAL_CFLAGS := $(WEBP_CFLAGS)
-LOCAL_C_INCLUDES := $(LOCAL_PATH)/../src
-LOCAL_STATIC_LIBRARIES := example_util imageio_util imagedec webp
+LOCAL_STATIC_LIBRARIES := example_util imageio_util imagedec webpdemux webp
 
 LOCAL_MODULE := cwebp
 
@@ -42,9 +41,7 @@ LOCAL_SRC_FILES := \
     dwebp.c \
 
 LOCAL_CFLAGS := $(WEBP_CFLAGS)
-LOCAL_C_INCLUDES := $(LOCAL_PATH)/../src
-LOCAL_STATIC_LIBRARIES := example_util imagedec imageenc webp
-
+LOCAL_STATIC_LIBRARIES := example_util imagedec imageenc webpdemux webp
 LOCAL_MODULE := dwebp
 
 include $(BUILD_EXECUTABLE)
@@ -58,7 +55,6 @@ LOCAL_SRC_FILES := \
     webpmux.c \
 
 LOCAL_CFLAGS := $(WEBP_CFLAGS)
-LOCAL_C_INCLUDES := $(LOCAL_PATH)/../src
 LOCAL_STATIC_LIBRARIES := example_util imageio_util webpmux webp
 
 LOCAL_MODULE := webpmux_example
@@ -74,8 +70,8 @@ LOCAL_SRC_FILES := \
     img2webp.c \
 
 LOCAL_CFLAGS := $(WEBP_CFLAGS)
-LOCAL_C_INCLUDES := $(LOCAL_PATH)/../src
-LOCAL_STATIC_LIBRARIES := example_util imageio_util imagedec webpmux webp
+LOCAL_STATIC_LIBRARIES := example_util imageio_util imagedec webpmux webpdemux \
+                          webp
 
 LOCAL_MODULE := img2webp_example
 
@@ -90,7 +86,6 @@ LOCAL_SRC_FILES := \
     webpinfo.c \
 
 LOCAL_CFLAGS := $(WEBP_CFLAGS)
-LOCAL_C_INCLUDES := $(LOCAL_PATH)/../src
 LOCAL_STATIC_LIBRARIES := example_util imageio_util webp
 
 LOCAL_MODULE := webpinfo_example

examples/Makefile.am

@@ -1,6 +1,9 @@
 AM_CPPFLAGS += -I$(top_builddir)/src -I$(top_srcdir)/src
 
-bin_PROGRAMS = dwebp cwebp
+bin_PROGRAMS =
+if BUILD_DEMUX
+  bin_PROGRAMS += dwebp cwebp
+endif
 if BUILD_ANIMDIFF
   noinst_PROGRAMS = anim_diff anim_dump
 endif
@@ -10,7 +13,7 @@ endif
 if BUILD_IMG2WEBP
   bin_PROGRAMS += img2webp
 endif
-if WANT_MUX
+if BUILD_MUX
   bin_PROGRAMS += webpmux
 endif
 if BUILD_VWEBP
@@ -25,16 +28,16 @@ noinst_LTLIBRARIES = libexample_util.la
 libexample_util_la_SOURCES = example_util.c example_util.h
 libexample_util_la_LIBADD = ../src/libwebp.la
 
-anim_diff_SOURCES = anim_diff.c anim_util.c anim_util.h
-anim_diff_CPPFLAGS = $(AM_CPPFLAGS) $(USE_EXPERIMENTAL_CODE) $(GIF_INCLUDES)
+anim_diff_SOURCES = anim_diff.c anim_util.c anim_util.h gifdec.c gifdec.h
+anim_diff_CPPFLAGS = $(AM_CPPFLAGS) $(GIF_INCLUDES)
 anim_diff_LDADD  =
 anim_diff_LDADD += ../src/demux/libwebpdemux.la
 anim_diff_LDADD += libexample_util.la
 anim_diff_LDADD += ../imageio/libimageio_util.la
 anim_diff_LDADD += $(GIF_LIBS) -lm
 
-anim_dump_SOURCES = anim_dump.c anim_util.c anim_util.h
-anim_dump_CPPFLAGS = $(AM_CPPFLAGS) $(USE_EXPERIMENTAL_CODE) $(PNG_INCLUDES)
+anim_dump_SOURCES = anim_dump.c anim_util.c anim_util.h gifdec.c gifdec.h
+anim_dump_CPPFLAGS = $(AM_CPPFLAGS) $(PNG_INCLUDES)
 anim_dump_CPPFLAGS += $(GIF_INCLUDES)
 anim_dump_LDADD  =
 anim_dump_LDADD += ../src/demux/libwebpdemux.la
@@ -44,7 +47,7 @@ anim_dump_LDADD += ../imageio/libimageenc.la
 anim_dump_LDADD += $(PNG_LIBS) $(GIF_LIBS) $(TIFF_LIBS) -lm
 
 cwebp_SOURCES  = cwebp.c stopwatch.h
-cwebp_CPPFLAGS  = $(AM_CPPFLAGS) $(USE_EXPERIMENTAL_CODE)
+cwebp_CPPFLAGS  = $(AM_CPPFLAGS)
 cwebp_LDADD  =
 cwebp_LDADD += libexample_util.la
 cwebp_LDADD += ../imageio/libimageio_util.la
@@ -53,7 +56,7 @@ cwebp_LDADD += ../src/libwebp.la
 cwebp_LDADD += $(JPEG_LIBS) $(PNG_LIBS) $(TIFF_LIBS)
 
 dwebp_SOURCES = dwebp.c stopwatch.h
-dwebp_CPPFLAGS  = $(AM_CPPFLAGS) $(USE_EXPERIMENTAL_CODE)
+dwebp_CPPFLAGS  = $(AM_CPPFLAGS)
 dwebp_CPPFLAGS += $(JPEG_INCLUDES) $(PNG_INCLUDES)
 dwebp_LDADD  =
 dwebp_LDADD += libexample_util.la
@@ -64,7 +67,7 @@ dwebp_LDADD += ../src/libwebp.la
 dwebp_LDADD +=$(PNG_LIBS) $(JPEG_LIBS)
 
 gif2webp_SOURCES = gif2webp.c gifdec.c gifdec.h
-gif2webp_CPPFLAGS = $(AM_CPPFLAGS) $(USE_EXPERIMENTAL_CODE) $(GIF_INCLUDES)
+gif2webp_CPPFLAGS = $(AM_CPPFLAGS) $(GIF_INCLUDES)
 gif2webp_LDADD  =
 gif2webp_LDADD += libexample_util.la
 gif2webp_LDADD += ../imageio/libimageio_util.la
@@ -73,7 +76,7 @@ gif2webp_LDADD += ../src/libwebp.la
 gif2webp_LDADD += $(GIF_LIBS)
 
 vwebp_SOURCES = vwebp.c
-vwebp_CPPFLAGS = $(AM_CPPFLAGS) $(USE_EXPERIMENTAL_CODE) $(GL_INCLUDES)
+vwebp_CPPFLAGS = $(AM_CPPFLAGS) $(GL_INCLUDES)
 vwebp_LDADD  =
 vwebp_LDADD += libexample_util.la
 vwebp_LDADD += ../imageio/libimageio_util.la
@@ -81,7 +84,7 @@ vwebp_LDADD += ../src/demux/libwebpdemux.la
 vwebp_LDADD += $(GL_LIBS)
 
 webpmux_SOURCES = webpmux.c
-webpmux_CPPFLAGS = $(AM_CPPFLAGS) $(USE_EXPERIMENTAL_CODE)
+webpmux_CPPFLAGS = $(AM_CPPFLAGS)
 webpmux_LDADD  =
 webpmux_LDADD += libexample_util.la
 webpmux_LDADD += ../imageio/libimageio_util.la
@@ -89,7 +92,7 @@ webpmux_LDADD += ../src/mux/libwebpmux.la
 webpmux_LDADD += ../src/libwebp.la
 
 img2webp_SOURCES = img2webp.c
-img2webp_CPPFLAGS = $(AM_CPPFLAGS) $(USE_EXPERIMENTAL_CODE)
+img2webp_CPPFLAGS = $(AM_CPPFLAGS)
 img2webp_LDADD  =
 img2webp_LDADD += libexample_util.la
 img2webp_LDADD += ../imageio/libimageio_util.la
@@ -99,7 +102,7 @@ img2webp_LDADD += ../src/libwebp.la
 img2webp_LDADD += $(PNG_LIBS) $(JPEG_LIBS) $(TIFF_LIBS)
 
 webpinfo_SOURCES = webpinfo.c
-webpinfo_CPPFLAGS = $(AM_CPPFLAGS) $(USE_EXPERIMENTAL_CODE)
+webpinfo_CPPFLAGS = $(AM_CPPFLAGS)
 webpinfo_LDADD  =
 webpinfo_LDADD += libexample_util.la
 webpinfo_LDADD += ../imageio/libimageio_util.la

examples/anim_diff.c

@@ -20,6 +20,7 @@
 #include <string.h>  // for 'strcmp'.
 
 #include "./anim_util.h"
+#include "./example_util.h"
 
 #if defined(_MSC_VER) && _MSC_VER < 1900
 #define snprintf _snprintf
@@ -197,11 +198,11 @@ static void Help(void) {
   printf("  -min_psnr <float> ... minimum per-frame PSNR\n");
   printf("  -raw_comparison ..... if this flag is not used, RGB is\n");
   printf("                        premultiplied before comparison\n");
-#ifdef WEBP_EXPERIMENTAL_FEATURES
-  printf("  -max_diff <int> ..... maximum allowed difference per channel "
-         "                        between corresponding pixels in subsequent"
+  printf("  -max_diff <int> ..... maximum allowed difference per channel\n"
+         "                        between corresponding pixels in subsequent\n"
          "                        frames\n");
-#endif
+  printf("  -h .................. this help\n");
+  printf("  -version ............ print version number and exit\n");
 }
 
 int main(int argc, const char* argv[]) {
@@ -217,11 +218,6 @@ int main(int argc, const char* argv[]) {
   const char* files[2] = { NULL, NULL };
   AnimatedImage images[2];
 
-  if (argc < 3) {
-    Help();
-    return -1;
-  }
-
   for (c = 1; c < argc; ++c) {
     int parse_error = 0;
     if (!strcmp(argv[c], "-dump_frames")) {
@@ -233,34 +229,30 @@ int main(int argc, const char* argv[]) {
       }
     } else if (!strcmp(argv[c], "-min_psnr")) {
       if (c < argc - 1) {
-        const char* const v = argv[++c];
-        char* end = NULL;
-        const double d = strtod(v, &end);
-        if (end == v) {
-          parse_error = 1;
-          fprintf(stderr, "Error! '%s' is not a floating point number.\n", v);
-        }
-        min_psnr = d;
+        min_psnr = ExUtilGetFloat(argv[++c], &parse_error);
       } else {
         parse_error = 1;
       }
     } else if (!strcmp(argv[c], "-raw_comparison")) {
       premultiply = 0;
-#ifdef WEBP_EXPERIMENTAL_FEATURES
     } else if (!strcmp(argv[c], "-max_diff")) {
       if (c < argc - 1) {
-        const char* const v = argv[++c];
-        char* end = NULL;
-        const int n = (int)strtol(v, &end, 10);
-        if (end == v) {
-          parse_error = 1;
-          fprintf(stderr, "Error! '%s' is not an integer.\n", v);
-        }
-        max_diff = n;
+        max_diff = ExUtilGetInt(argv[++c], 0, &parse_error);
       } else {
         parse_error = 1;
       }
-#endif
+    } else if (!strcmp(argv[c], "-h") || !strcmp(argv[c], "-help")) {
+      Help();
+      return 0;
+    } else if (!strcmp(argv[c], "-version")) {
+      int dec_version, demux_version;
+      GetAnimatedImageVersions(&dec_version, &demux_version);
+      printf("WebP Decoder version: %d.%d.%d\nWebP Demux version: %d.%d.%d\n",
+             (dec_version >> 16) & 0xff, (dec_version >> 8) & 0xff,
+             (dec_version >> 0) & 0xff,
+             (demux_version >> 16) & 0xff, (demux_version >> 8) & 0xff,
+             (demux_version >> 0) & 0xff);
+      return 0;
     } else {
       if (!got_input1) {
         files[0] = argv[c];
@@ -277,6 +269,12 @@ int main(int argc, const char* argv[]) {
       return -1;
     }
   }
+  if (argc < 3) {
+    Help();
+    return -1;
+  }
+
+
   if (!got_input2) {
     Help();
     return -1;

examples/anim_dump.c

@@ -30,6 +30,8 @@ static void Help(void) {
                                   "(default: 'dump_')\n");
   printf("  -tiff ............... save frames as TIFF\n");
   printf("  -pam ................ save frames as PAM\n");
+  printf("  -h .................. this help\n");
+  printf("  -version ............ print version number and exit\n");
 }
 
 int main(int argc, const char* argv[]) {
@@ -66,6 +68,18 @@ int main(int argc, const char* argv[]) {
     } else if (!strcmp(argv[c], "-pam")) {
       format = PAM;
       suffix = "pam";
+    } else if (!strcmp(argv[c], "-h") || !strcmp(argv[c], "-help")) {
+      Help();
+      return 0;
+    } else if (!strcmp(argv[c], "-version")) {
+      int dec_version, demux_version;
+      GetAnimatedImageVersions(&dec_version, &demux_version);
+      printf("WebP Decoder version: %d.%d.%d\nWebP Demux version: %d.%d.%d\n",
+             (dec_version >> 16) & 0xff, (dec_version >> 8) & 0xff,
+             (dec_version >> 0) & 0xff,
+             (demux_version >> 16) & 0xff, (demux_version >> 8) & 0xff,
+             (demux_version >> 0) & 0xff);
+      return 0;
     } else {
       uint32_t i;
       AnimatedImage image;

examples/anim_util.c

@@ -23,6 +23,7 @@
 #include "webp/decode.h"
 #include "webp/demux.h"
 #include "../imageio/imageio_util.h"
+#include "./gifdec.h"
 
 #if defined(_MSC_VER) && _MSC_VER < 1900
 #define snprintf _snprintf
@@ -370,26 +371,6 @@ static int DGifSavedExtensionToGCB(GifFileType* GifFile, int ImageIndex,
 #define DGifCloseFile(a, b) DGifCloseFile(a)
 #endif
 
-static void GIFDisplayError(const GifFileType* const gif, int gif_error) {
-  // libgif 4.2.0 has retired PrintGifError() and added GifErrorString().
-#if LOCAL_GIF_PREREQ(4, 2)
-#if LOCAL_GIF_PREREQ(5, 0)
-  const char* error_str =
-      GifErrorString((gif == NULL) ? gif_error : gif->Error);
-#else
-  const char* error_str = GifErrorString();
-  (void)gif;
-#endif
-  if (error_str == NULL) error_str = "Unknown error";
-  fprintf(stderr, "GIFLib Error %d: %s\n", gif_error, error_str);
-#else
-  (void)gif;
-  fprintf(stderr, "GIFLib Error %d: ", gif_error);
-  PrintGifError();
-  fprintf(stderr, "\n");
-#endif
-}
-
 static int IsKeyFrameGIF(const GifImageDesc* prev_desc, int prev_dispose,
                          const DecodedFrame* const prev_frame,
                          int canvas_width, int canvas_height) {
@@ -594,6 +575,9 @@ static int ReadAnimatedGIF(const char filename[], AnimatedImage* const image,
     curr_frame = &image->frames[i];
     curr_rgba = curr_frame->rgba;
     curr_frame->duration = GetFrameDurationGIF(gif, i);
+    // Force frames with a small or no duration to 100ms to be consistent
+    // with web browsers and other transcoding tools (like gif2webp itself).
+    if (curr_frame->duration <= 10) curr_frame->duration = 100;
 
     if (i == 0) {  // Initialize as transparent.
       curr_frame->is_key_frame = 1;
@@ -785,3 +769,9 @@ void GetDiffAndPSNR(const uint8_t rgba1[], const uint8_t rgba2[],
     *psnr = 4.3429448 * log(255. * 255. / sse);
   }
 }
+
+void GetAnimatedImageVersions(int* const decoder_version,
+                              int* const demux_version) {
+  *decoder_version = WebPGetDecoderVersion();
+  *demux_version = WebPGetDemuxVersion();
+}

examples/anim_util.h

@@ -62,6 +62,10 @@ void GetDiffAndPSNR(const uint8_t rgba1[], const uint8_t rgba2[],
                     uint32_t width, uint32_t height, int premultiply,
                     int* const max_diff, double* const psnr);
 
+// Return library versions used by anim_util.
+void GetAnimatedImageVersions(int* const decoder_version,
+                              int* const demux_version);
+
 #ifdef __cplusplus
 }    // extern "C"
 #endif

examples/cwebp.c

@@ -140,10 +140,11 @@ static void PrintByteCount(const int bytes[4], int total_size,
   fprintf(stderr, "| %7d  (%.1f%%)\n", total, 100.f * total / total_size);
 }
 
-static void PrintPercents(const int counts[4], int total) {
+static void PrintPercents(const int counts[4]) {
   int s;
+  const int total = counts[0] + counts[1] + counts[2] + counts[3];
   for (s = 0; s < 4; ++s) {
-    fprintf(stderr, "|      %2d%%", 100 * counts[s] / total);
+    fprintf(stderr, "|      %2d%%", (int)(100. * counts[s] / total + .5));
   }
   fprintf(stderr, "| %7d\n", total);
 }
@@ -186,7 +187,8 @@ static void PrintExtraInfoLossless(const WebPPicture* const pic,
   } else {
     fprintf(stderr, "File:      %s\n", file_name);
     fprintf(stderr, "Dimension: %d x %d\n", pic->width, pic->height);
-    fprintf(stderr, "Output:    %d bytes\n", stats->coded_size);
+    fprintf(stderr, "Output:    %d bytes (%.2f bpp)\n", stats->coded_size,
+            8.f * stats->coded_size / pic->width / pic->height);
     PrintFullLosslessInfo(stats, "ARGB");
   }
 }
@@ -207,15 +209,18 @@ static void PrintExtraInfoLossy(const WebPPicture* const pic, int short_output,
             pic->width, pic->height,
             stats->alpha_data_size ? " (with alpha)" : "");
     fprintf(stderr, "Output:    "
-            "%d bytes Y-U-V-All-PSNR %2.2f %2.2f %2.2f   %2.2f dB\n",
+            "%d bytes Y-U-V-All-PSNR %2.2f %2.2f %2.2f   %2.2f dB\n"
+            "           (%.2f bpp)\n",
             stats->coded_size,
-            stats->PSNR[0], stats->PSNR[1], stats->PSNR[2], stats->PSNR[3]);
+            stats->PSNR[0], stats->PSNR[1], stats->PSNR[2], stats->PSNR[3],
+            8.f * stats->coded_size / pic->width / pic->height);
     if (total > 0) {
       int totals[4] = { 0, 0, 0, 0 };
-      fprintf(stderr, "block count:  intra4: %d\n"
-                      "              intra16: %d  (-> %.2f%%)\n",
-              num_i4, num_i16, 100.f * num_i16 / total);
-      fprintf(stderr, "              skipped block: %d (%.2f%%)\n",
+      fprintf(stderr, "block count:  intra4:     %6d  (%.2f%%)\n"
+                      "              intra16:    %6d  (%.2f%%)\n"
+                      "              skipped:    %6d  (%.2f%%)\n",
+              num_i4, 100.f * num_i4 / total,
+              num_i16, 100.f * num_i16 / total,
               num_skip, 100.f * num_skip / total);
       fprintf(stderr, "bytes used:  header:         %6d  (%.1f%%)\n"
                       "             mode-partition: %6d  (%.1f%%)\n",
@@ -239,7 +244,7 @@ static void PrintExtraInfoLossy(const WebPPicture* const pic, int short_output,
         PrintByteCount(stats->residual_bytes[2], stats->coded_size, totals);
       }
       fprintf(stderr, "    macroblocks:  ");
-      PrintPercents(stats->segment_size, total);
+      PrintPercents(stats->segment_size);
       fprintf(stderr, "      quantizer:  ");
       PrintValues(stats->segment_quant);
       fprintf(stderr, "   filter level:  ");
@@ -580,9 +585,6 @@ static void HelpLong(void) {
   printf("  -near_lossless <int> ... use near-lossless image\n"
          "                           preprocessing (0..100=off), "
          "default=100\n");
-#ifdef WEBP_EXPERIMENTAL_FEATURES  /* not documented yet */
-  printf("  -delta_palette ......... use delta palettization\n");
-#endif  // WEBP_EXPERIMENTAL_FEATURES
   printf("  -hint <string> ......... specify image characteristics hint,\n");
   printf("                           one of: photo, picture or graph\n");
 
@@ -751,11 +753,6 @@ int main(int argc, const char *argv[]) {
     } else if (!strcmp(argv[c], "-near_lossless") && c < argc - 1) {
       config.near_lossless = ExUtilGetInt(argv[++c], 0, &parse_error);
       config.lossless = 1;  // use near-lossless only with lossless
-#ifdef WEBP_EXPERIMENTAL_FEATURES
-    } else if (!strcmp(argv[c], "-delta_palette")) {
-      config.use_delta_palette = 1;
-      config.lossless = 1;  // delta-palette is for lossless only
-#endif  // WEBP_EXPERIMENTAL_FEATURES
     } else if (!strcmp(argv[c], "-hint") && c < argc - 1) {
       ++c;
       if (!strcmp(argv[c], "photo")) {
@@ -1015,10 +1012,53 @@ int main(int argc, const char *argv[]) {
     }
   }
   if ((resize_w | resize_h) > 0) {
+    WebPPicture picture_no_alpha;
+    if (config.exact) {
+      // If -exact, we can't premultiply RGB by A otherwise RGB is lost if A=0.
+      // We rescale an opaque copy and assemble scaled A and non-premultiplied
+      // RGB channels. This is slower but it's a very uncommon use case. Color
+      // leak at sharp alpha edges is possible.
+      if (!WebPPictureCopy(&picture, &picture_no_alpha)) {
+        fprintf(stderr, "Error! Cannot copy temporary picture\n");
+        goto Error;
+      }
+
+      // We enforced picture.use_argb = 1 above. Now, remove the alpha values.
+      {
+        int x, y;
+        uint32_t* argb_no_alpha = picture_no_alpha.argb;
+        for (y = 0; y < picture_no_alpha.height; ++y) {
+          for (x = 0; x < picture_no_alpha.width; ++x) {
+            argb_no_alpha[x] |= 0xff000000;  // Opaque copy.
+          }
+          argb_no_alpha += picture_no_alpha.argb_stride;
+        }
+      }
+
+      if (!WebPPictureRescale(&picture_no_alpha, resize_w, resize_h)) {
+        fprintf(stderr, "Error! Cannot resize temporary picture\n");
+        goto Error;
+      }
+    }
+
     if (!WebPPictureRescale(&picture, resize_w, resize_h)) {
       fprintf(stderr, "Error! Cannot resize picture\n");
       goto Error;
     }
+
+    if (config.exact) {  // Put back the alpha information.
+      int x, y;
+      uint32_t* argb_no_alpha = picture_no_alpha.argb;
+      uint32_t* argb = picture.argb;
+      for (y = 0; y < picture_no_alpha.height; ++y) {
+        for (x = 0; x < picture_no_alpha.width; ++x) {
+          argb[x] = (argb[x] & 0xff000000) | (argb_no_alpha[x] & 0x00ffffff);
+        }
+        argb_no_alpha += picture_no_alpha.argb_stride;
+        argb += picture.argb_stride;
+      }
+      WebPPictureFree(&picture_no_alpha);
+    }
   }
   if (verbose && (crop != 0 || (resize_w | resize_h) > 0)) {
     const double preproc_time = StopwatchReadAndReset(&stop_watch);
@@ -1050,7 +1090,8 @@ int main(int argc, const char *argv[]) {
   // Write info
   if (dump_file) {
     if (picture.use_argb) {
-      fprintf(stderr, "Warning: can't dump file (-d option) in lossless mode.");
+      fprintf(stderr, "Warning: can't dump file (-d option) "
+                      "in lossless mode.\n");
     } else if (!DumpPicture(&picture, dump_file)) {
       fprintf(stderr, "Warning, couldn't dump picture %s\n", dump_file);
     }

examples/example_util.c

@@ -12,10 +12,14 @@
 
 #include "./example_util.h"
 
+#include <assert.h>
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
 
+#include "webp/mux_types.h"
+#include "../imageio/imageio_util.h"
+
 //------------------------------------------------------------------------------
 // String parsing
 
@@ -56,3 +60,68 @@ float ExUtilGetFloat(const char* const v, int* const error) {
   }
   return f;
 }
+
+//------------------------------------------------------------------------------
+
+static void ResetCommandLineArguments(int argc, const char* argv[],
+                                      CommandLineArguments* const args) {
+  assert(args != NULL);
+  args->argc_ = argc;
+  args->argv_ = argv;
+  args->own_argv_ = 0;
+  WebPDataInit(&args->argv_data_);
+}
+
+void ExUtilDeleteCommandLineArguments(CommandLineArguments* const args) {
+  if (args != NULL) {
+    if (args->own_argv_) {
+      free((void*)args->argv_);
+      WebPDataClear(&args->argv_data_);
+    }
+    ResetCommandLineArguments(0, NULL, args);
+  }
+}
+
+#define MAX_ARGC 16384
+int ExUtilInitCommandLineArguments(int argc, const char* argv[],
+                                   CommandLineArguments* const args) {
+  if (args == NULL || argv == NULL) return 0;
+  ResetCommandLineArguments(argc, argv, args);
+  if (argc == 1 && argv[0][0] != '-') {
+    char* cur;
+    const char sep[] = " \t\r\n\f\v";
+    if (!ExUtilReadFileToWebPData(argv[0], &args->argv_data_)) {
+      return 0;
+    }
+    args->own_argv_ = 1;
+    args->argv_ = (const char**)malloc(MAX_ARGC * sizeof(*args->argv_));
+    if (args->argv_ == NULL) return 0;
+
+    argc = 0;
+    for (cur = strtok((char*)args->argv_data_.bytes, sep);
+         cur != NULL;
+         cur = strtok(NULL, sep)) {
+      if (argc == MAX_ARGC) {
+        fprintf(stderr, "ERROR: Arguments limit %d reached\n", MAX_ARGC);
+        return 0;
+      }
+      assert(strlen(cur) != 0);
+      args->argv_[argc++] = cur;
+    }
+    args->argc_ = argc;
+  }
+  return 1;
+}
+
+//------------------------------------------------------------------------------
+
+int ExUtilReadFileToWebPData(const char* const filename,
+                             WebPData* const webp_data) {
+  const uint8_t* data;
+  size_t size;
+  if (webp_data == NULL) return 0;
+  if (!ImgIoUtilReadFile(filename, &data, &size)) return 0;
+  webp_data->bytes = data;
+  webp_data->size = size;
+  return 1;
+}

examples/example_util.h

@@ -14,6 +14,7 @@
 #define WEBP_EXAMPLES_EXAMPLE_UTIL_H_
 
 #include "webp/types.h"
+#include "webp/mux_types.h"
 
 #ifdef __cplusplus
 extern "C" {
@@ -35,6 +36,33 @@ float ExUtilGetFloat(const char* const v, int* const error);
 // actually parsed is returned, or -1 if an error occurred.
 int ExUtilGetInts(const char* v, int base, int max_output, int output[]);
 
+// Reads a file named 'filename' into a WebPData structure. The content of
+// webp_data is overwritten. Returns false in case of error.
+int ExUtilReadFileToWebPData(const char* const filename,
+                             WebPData* const webp_data);
+
+//------------------------------------------------------------------------------
+// Command-line arguments
+
+typedef struct {
+  int argc_;
+  const char** argv_;
+  WebPData argv_data_;
+  int own_argv_;
+} CommandLineArguments;
+
+// Initializes the structure from the command-line parameters. If there is
+// only one parameter and it does not start with a '-', then it is assumed to
+// be a file name. This file will be read and tokenized into command-line
+// arguments. The content of 'args' is overwritten.
+// Returns false in case of error (memory allocation failure, non
+// existing file, too many arguments, ...).
+int ExUtilInitCommandLineArguments(int argc, const char* argv[],
+                                   CommandLineArguments* const args);
+
+// Deallocate all memory and reset 'args'.
+void ExUtilDeleteCommandLineArguments(CommandLineArguments* const args);
+
 #ifdef __cplusplus
 }    // extern "C"
 #endif

examples/gif2webp.c

@@ -23,6 +23,10 @@
 
 #ifdef WEBP_HAVE_GIF
 
+#if defined(HAVE_UNISTD_H) && HAVE_UNISTD_H
+#include <unistd.h>
+#endif
+
 #include <gif_lib.h>
 #include "webp/encode.h"
 #include "webp/mux.h"
@@ -30,6 +34,10 @@
 #include "../imageio/imageio_util.h"
 #include "./gifdec.h"
 
+#if !defined(STDIN_FILENO)
+#define STDIN_FILENO 0
+#endif
+
 //------------------------------------------------------------------------------
 
 static int transparent_index = GIF_INDEX_INVALID;  // Opaque by default.
@@ -92,7 +100,6 @@ int main(int argc, const char *argv[]) {
   WebPMuxError err = WEBP_MUX_OK;
   int ok = 0;
   const char *in_file = NULL, *out_file = NULL;
-  FILE* out = NULL;
   GifFileType* gif = NULL;
   int frame_duration = 0;
   int frame_timestamp = 0;
@@ -263,9 +270,11 @@ int main(int argc, const char *argv[]) {
 
   // Start the decoder object
 #if LOCAL_GIF_PREREQ(5,0)
-  gif = DGifOpenFileName(in_file, &gif_error);
+  gif = !strcmp(in_file, "-") ? DGifOpenFileHandle(STDIN_FILENO, &gif_error)
+                              : DGifOpenFileName(in_file, &gif_error);
 #else
-  gif = DGifOpenFileName(in_file);
+  gif = !strcmp(in_file, "-") ? DGifOpenFileHandle(STDIN_FILENO)
+                              : DGifOpenFileName(in_file);
 #endif
   if (gif == NULL) goto End;
 
@@ -351,6 +360,14 @@ int main(int argc, const char *argv[]) {
         GIFDisposeFrame(orig_dispose, &gif_rect, &prev_canvas, &curr_canvas);
         GIFCopyPixels(&curr_canvas, &prev_canvas);
 
+        // Force frames with a small or no duration to 100ms to be consistent
+        // with web browsers and other transcoding tools. This also avoids
+        // incorrect durations between frames when padding frames are
+        // discarded.
+        if (frame_duration <= 10) {
+          frame_duration = 100;
+        }
+
         // Update timestamp (for next frame).
         frame_timestamp += frame_duration;
 
@@ -532,8 +549,13 @@ int main(int argc, const char *argv[]) {
       goto End;
     }
     if (!quiet) {
-      fprintf(stderr, "Saved output file (%d bytes): %s\n",
-              (int)webp_data.size, out_file);
+      if (!strcmp(out_file, "-")) {
+        fprintf(stderr, "Saved %d bytes to STDIO\n",
+                (int)webp_data.size);
+      } else {
+        fprintf(stderr, "Saved output file (%d bytes): %s\n",
+                (int)webp_data.size, out_file);
+      }
     }
   } else {
     if (!quiet) {
@@ -555,7 +577,6 @@ int main(int argc, const char *argv[]) {
   WebPPictureFree(&curr_canvas);
   WebPPictureFree(&prev_canvas);
   WebPAnimEncoderDelete(enc);
-  if (out != NULL && out_file != NULL) fclose(out);
 
   if (gif_error != GIF_OK) {
     GIFDisplayError(gif, gif_error);

examples/img2webp.c

@@ -48,6 +48,7 @@ static void Help(void) {
   printf(" -mixed ............... use mixed lossy/lossless automatic mode\n");
   printf(" -v ................... verbose mode\n");
   printf(" -h ................... this help\n");
+  printf(" -version ............. print version number and exit\n");
   printf("\n");
 
   printf("Per-frame options (only used for subsequent images input):\n");
@@ -60,6 +61,10 @@ static void Help(void) {
   printf("\n");
   printf("example: img2webp -loop 2 in0.png -lossy in1.jpg\n"
          "                  -d 80 in2.tiff -o out.webp\n");
+  printf("\nNote: if a single file name is passed as the argument, the "
+         "arguments will be\n");
+  printf("tokenized from this file. The file name must not start with "
+         "the character '-'.\n");
 }
 
 //------------------------------------------------------------------------------
@@ -117,14 +122,13 @@ static int SetLoopCount(int loop_count, WebPData* const webp_data) {
 
 //------------------------------------------------------------------------------
 
-int main(int argc, char* argv[]) {
+int main(int argc, const char* argv[]) {
   const char* output = NULL;
   WebPAnimEncoder* enc = NULL;
   int verbose = 0;
   int pic_num = 0;
   int duration = 100;
   int timestamp_ms = 0;
-  int ok = 1;
   int loop_count = 0;
   int width = 0, height = 0;
   WebPAnimEncoderOptions anim_config;
@@ -133,17 +137,23 @@ int main(int argc, char* argv[]) {
   WebPData webp_data;
   int c;
   int have_input = 0;
+  CommandLineArguments cmd_args;
+  int ok = ExUtilInitCommandLineArguments(argc - 1, argv + 1, &cmd_args);
+  if (!ok) return 1;
+  argc = cmd_args.argc_;
+  argv = cmd_args.argv_;
 
   WebPDataInit(&webp_data);
   if (!WebPAnimEncoderOptionsInit(&anim_config) ||
       !WebPConfigInit(&config) ||
       !WebPPictureInit(&pic)) {
     fprintf(stderr, "Library version mismatch!\n");
-    return 1;
+    ok = 0;
+    goto End;
   }
 
   // 1st pass of option parsing
-  for (c = 1; ok && c < argc; ++c) {
+  for (c = 0; ok && c < argc; ++c) {
     if (argv[c][0] == '-') {
       int parse_error = 0;
       if (!strcmp(argv[c], "-o") && c + 1 < argc) {
@@ -171,7 +181,15 @@ int main(int argc, char* argv[]) {
         verbose = 1;
       } else if (!strcmp(argv[c], "-h") || !strcmp(argv[c], "-help")) {
         Help();
-        return 0;
+        goto End;
+      } else if (!strcmp(argv[c], "-version")) {
+        const int enc_version = WebPGetEncoderVersion();
+        const int mux_version = WebPGetMuxVersion();
+        printf("WebP Encoder version: %d.%d.%d\nWebP Mux version: %d.%d.%d\n",
+               (enc_version >> 16) & 0xff, (enc_version >> 8) & 0xff,
+               enc_version & 0xff, (mux_version >> 16) & 0xff,
+               (mux_version >> 8) & 0xff, mux_version & 0xff);
+        goto End;
       } else {
         continue;
       }
@@ -184,13 +202,13 @@ int main(int argc, char* argv[]) {
   }
   if (!have_input) {
     fprintf(stderr, "No input file(s) for generating animation!\n");
-    return 0;
+    goto End;
   }
 
   // image-reading pass
   pic_num = 0;
   config.lossless = 1;
-  for (c = 1; ok && c < argc; ++c) {
+  for (c = 0; ok && c < argc; ++c) {
     if (argv[c] == NULL) continue;
     if (argv[c][0] == '-') {    // parse local options
       int parse_error = 0;
@@ -294,7 +312,7 @@ int main(int argc, char* argv[]) {
     fprintf(stderr, "[%d frames, %u bytes].\n",
             pic_num, (unsigned int)webp_data.size);
   }
-
   WebPDataClear(&webp_data);
+  ExUtilDeleteCommandLineArguments(&cmd_args);
   return ok ? 0 : 1;
 }

examples/stopwatch.h

@@ -60,4 +60,4 @@ static WEBP_INLINE double StopwatchReadAndReset(Stopwatch* watch) {
 
 #endif   /* _WIN32 */
 
-#endif  /* WEBP_EXAMPLES_STOPWATCH_H_ */
+#endif  // WEBP_EXAMPLES_STOPWATCH_H_

examples/vwebp.c

@@ -56,6 +56,7 @@ static struct {
   int print_info;
   int only_deltas;
   int use_color_profile;
+  int draw_anim_background_color;
 
   int canvas_width, canvas_height;
   int loop_count;
@@ -205,6 +206,11 @@ static void decode_callback(int what) {
         }
       }
       duration = curr->duration;
+      // Behavior copied from Chrome, cf:
+      // https://cs.chromium.org/chromium/src/third_party/WebKit/Source/
+      // platform/graphics/DeferredImageDecoder.cpp?
+      // rcl=b4c33049f096cd283f32be9a58b9a9e768227c26&l=246
+      if (duration <= 10) duration = 100;
     }
     if (!Decode()) {
       kParams.decoding_error = 1;
@@ -220,6 +226,9 @@ static void decode_callback(int what) {
 // Callbacks
 
 static void HandleKey(unsigned char key, int pos_x, int pos_y) {
+  // Note: rescaling the window or toggling some features during an animation
+  // generates visual artifacts. This is not fixed because refreshing the frame
+  // may require rendering the whole animation from start till current frame.
   (void)pos_x;
   (void)pos_y;
   if (key == 'q' || key == 'Q' || key == 27 /* Esc */) {
@@ -247,9 +256,11 @@ static void HandleKey(unsigned char key, int pos_x, int pos_y) {
         glutPostRedisplay();
       }
     }
+  } else if (key == 'b') {
+    kParams.draw_anim_background_color = 1 - kParams.draw_anim_background_color;
+    if (!kParams.has_animation) ClearPreviousFrame();
+    glutPostRedisplay();
   } else if (key == 'i') {
-    // Note: doesn't handle refresh of animation's last-frame (it's quite
-    // more involved to do, since you need to save the previous frame).
     kParams.print_info = 1 - kParams.print_info;
     if (!kParams.has_animation) ClearPreviousFrame();
     glutPostRedisplay();
@@ -281,7 +292,7 @@ static void PrintString(const char* const text) {
 }
 
 static float GetColorf(uint32_t color, int shift) {
-  return (color >> shift) / 255.f;
+  return ((color >> shift) & 0xff) / 255.f;
 }
 
 static void DrawCheckerBoard(void) {
@@ -304,6 +315,43 @@ static void DrawCheckerBoard(void) {
   glPopMatrix();
 }
 
+static void DrawBackground(void) {
+  // Whole window cleared with clear color, checkerboard rendered on top of it.
+  glClear(GL_COLOR_BUFFER_BIT);
+  DrawCheckerBoard();
+
+  // ANIM background color rendered (blend) on top. Default is white for still
+  // images (without ANIM chunk). glClear() can't be used for that (no blend).
+  if (kParams.draw_anim_background_color) {
+    glPushMatrix();
+    glLoadIdentity();
+    glColor4f(GetColorf(kParams.bg_color, 16),  // BGRA from spec
+              GetColorf(kParams.bg_color, 8),
+              GetColorf(kParams.bg_color, 0),
+              GetColorf(kParams.bg_color, 24));
+    glRecti(-1, -1, +1, +1);
+    glPopMatrix();
+  }
+}
+
+// Draw background in a scissored rectangle.
+static void DrawBackgroundScissored(int window_x, int window_y, int frame_w,
+                                    int frame_h) {
+  // Only update the requested area, not the whole canvas.
+  window_x = window_x * kParams.viewport_width / kParams.canvas_width;
+  window_y = window_y * kParams.viewport_height / kParams.canvas_height;
+  frame_w = frame_w * kParams.viewport_width / kParams.canvas_width;
+  frame_h = frame_h * kParams.viewport_height / kParams.canvas_height;
+
+  // glScissor() takes window coordinates (0,0 at bottom left).
+  window_y = kParams.viewport_height - window_y - frame_h;
+
+  glEnable(GL_SCISSOR_TEST);
+  glScissor(window_x, window_y, frame_w, frame_h);
+  DrawBackground();
+  glDisable(GL_SCISSOR_TEST);
+}
+
 static void HandleDisplay(void) {
   const WebPDecBuffer* const pic = kParams.pic;
   const WebPIterator* const curr = &kParams.curr_frame;
@@ -320,38 +368,21 @@ static void HandleDisplay(void) {
   glPixelStorei(GL_UNPACK_ROW_LENGTH, pic->u.RGBA.stride / 4);
 
   if (kParams.only_deltas) {
-    DrawCheckerBoard();
-  } else if (prev->dispose_method == WEBP_MUX_DISPOSE_BACKGROUND ||
-      curr->blend_method == WEBP_MUX_NO_BLEND) {
-    // glScissor() takes window coordinates (0,0 at bottom left).
-    int window_x, window_y;
-    int frame_w, frame_h;
+    DrawBackground();
+  } else {
+    // The rectangle of the previous frame might be different than the current
+    // frame, so we may need to DrawBackgroundScissored for both.
     if (prev->dispose_method == WEBP_MUX_DISPOSE_BACKGROUND) {
       // Clear the previous frame rectangle.
-      window_x = prev->x_offset;
-      window_y = kParams.canvas_height - prev->y_offset - prev->height;
-      frame_w = prev->width;
-      frame_h = prev->height;
-    } else {  // curr->blend_method == WEBP_MUX_NO_BLEND.
-      // We simulate no-blending behavior by first clearing the current frame
-      // rectangle (to a checker-board) and then alpha-blending against it.
-      window_x = curr->x_offset;
-      window_y = kParams.canvas_height - curr->y_offset - curr->height;
-      frame_w = curr->width;
-      frame_h = curr->height;
+      DrawBackgroundScissored(prev->x_offset, prev->y_offset, prev->width,
+                              prev->height);
+    }
+    if (curr->blend_method == WEBP_MUX_NO_BLEND) {
+      // We simulate no-blending behavior by first clearing the current frame
+      // rectangle and then alpha-blending against it.
+      DrawBackgroundScissored(curr->x_offset, curr->y_offset, curr->width,
+                              curr->height);
     }
-    glEnable(GL_SCISSOR_TEST);
-    // Only update the requested area, not the whole canvas.
-    window_x = window_x * kParams.viewport_width / kParams.canvas_width;
-    window_y = window_y * kParams.viewport_height / kParams.canvas_height;
-    frame_w = frame_w * kParams.viewport_width / kParams.canvas_width;
-    frame_h = frame_h * kParams.viewport_height / kParams.canvas_height;
-    glScissor(window_x, window_y, frame_w, frame_h);
-
-    glClear(GL_COLOR_BUFFER_BIT);  // use clear color
-    DrawCheckerBoard();
-
-    glDisable(GL_SCISSOR_TEST);
   }
 
   *prev = *curr;
@@ -403,37 +434,35 @@ static void StartDisplay(void) {
   glutKeyboardFunc(HandleKey);
   glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
   glEnable(GL_BLEND);
-  glClearColor(GetColorf(kParams.bg_color, 0),
-               GetColorf(kParams.bg_color, 8),
-               GetColorf(kParams.bg_color, 16),
-               GetColorf(kParams.bg_color, 24));
-  glClear(GL_COLOR_BUFFER_BIT);
-  DrawCheckerBoard();
+  glClearColor(0, 0, 0, 0);  // window will be cleared to black (no blend)
+  DrawBackground();
 }
 
 //------------------------------------------------------------------------------
 // Main
 
 static void Help(void) {
-  printf("Usage: vwebp in_file [options]\n\n"
-         "Decodes the WebP image file and visualize it using OpenGL\n"
-         "Options are:\n"
-         "  -version ..... print version number and exit\n"
-         "  -noicc ....... don't use the icc profile if present\n"
-         "  -nofancy ..... don't use the fancy YUV420 upscaler\n"
-         "  -nofilter .... disable in-loop filtering\n"
-         "  -dither <int>  dithering strength (0..100), default=50\n"
-         "  -noalphadither disable alpha plane dithering\n"
-         "  -mt .......... use multi-threading\n"
-         "  -info ........ print info\n"
-         "  -h ........... this help message\n"
-         "\n"
-         "Keyboard shortcuts:\n"
-         "  'c' ................ toggle use of color profile\n"
-         "  'i' ................ overlay file information\n"
-         "  'd' ................ disable blending & disposal (debug)\n"
-         "  'q' / 'Q' / ESC .... quit\n"
-        );
+  printf(
+      "Usage: vwebp in_file [options]\n\n"
+      "Decodes the WebP image file and visualize it using OpenGL\n"
+      "Options are:\n"
+      "  -version ..... print version number and exit\n"
+      "  -noicc ....... don't use the icc profile if present\n"
+      "  -nofancy ..... don't use the fancy YUV420 upscaler\n"
+      "  -nofilter .... disable in-loop filtering\n"
+      "  -dither <int>  dithering strength (0..100), default=50\n"
+      "  -noalphadither disable alpha plane dithering\n"
+      "  -usebgcolor .. display background color\n"
+      "  -mt .......... use multi-threading\n"
+      "  -info ........ print info\n"
+      "  -h ........... this help message\n"
+      "\n"
+      "Keyboard shortcuts:\n"
+      "  'c' ................ toggle use of color profile\n"
+      "  'b' ................ toggle background color display\n"
+      "  'i' ................ overlay file information\n"
+      "  'd' ................ disable blending & disposal (debug)\n"
+      "  'q' / 'Q' / ESC .... quit\n");
 }
 
 int main(int argc, char *argv[]) {
@@ -448,6 +477,8 @@ int main(int argc, char *argv[]) {
   config->options.dithering_strength = 50;
   config->options.alpha_dithering_strength = 100;
   kParams.use_color_profile = 1;
+  // Background color hidden by default to see transparent areas.
+  kParams.draw_anim_background_color = 0;
 
   for (c = 1; c < argc; ++c) {
     int parse_error = 0;
@@ -462,6 +493,8 @@ int main(int argc, char *argv[]) {
       config->options.bypass_filtering = 1;
     } else if (!strcmp(argv[c], "-noalphadither")) {
       config->options.alpha_dithering_strength = 0;
+    } else if (!strcmp(argv[c], "-usebgcolor")) {
+      kParams.draw_anim_background_color = 1;
     } else if (!strcmp(argv[c], "-dither") && c + 1 < argc) {
       config->options.dithering_strength =
           ExUtilGetInt(argv[++c], 0, &parse_error);

examples/webpinfo.c

@@ -340,7 +340,7 @@ static WebPInfoStatus ParseLossyHeader(const ChunkData* const chunk_data,
   WebPInfoStatus status = WEBP_INFO_OK;
   uint64_t bit_position = 0;
   uint64_t* const bit_pos = &bit_position;
-  int color_space, clamp_type;
+  int colorspace, clamp_type;
   printf("  Parsing lossy bitstream...\n");
   // Calling WebPGetFeatures() in ProcessImageChunk() should ensure this.
   assert(chunk_data->size_ >= CHUNK_HEADER_SIZE + 10);
@@ -381,9 +381,9 @@ static WebPInfoStatus ParseLossyHeader(const ChunkData* const chunk_data,
     LOG_ERROR("Bad partition length.");
     return WEBP_INFO_BITSTREAM_ERROR;
   }
-  GET_BITS(color_space, 1);
+  GET_BITS(colorspace, 1);
   GET_BITS(clamp_type, 1);
-  printf("  Color space:      %d\n", color_space);
+  printf("  Color space:      %d\n", colorspace);
   printf("  Clamp type:       %d\n", clamp_type);
   status = ParseLossySegmentHeader(webp_info, data, data_size, bit_pos);
   if (status != WEBP_INFO_OK) return status;

examples/webpmux.c

@@ -47,6 +47,7 @@
     webpmux -info in.webp
     webpmux [ -h | -help ]
     webpmux -version
+    webpmux argument_file_name
 */
 
 #ifdef HAVE_CONFIG_H
@@ -108,28 +109,26 @@ static const char* const kDescriptions[LAST_FEATURE] = {
 };
 
 typedef struct {
-  FeatureType type_;
-  FeatureArg* args_;
-  int arg_count_;
-} Feature;
+  CommandLineArguments cmd_args_;
 
-typedef struct {
   ActionType action_type_;
   const char* input_;
   const char* output_;
-  Feature feature_;
-} WebPMuxConfig;
+  FeatureType type_;
+  FeatureArg* args_;
+  int arg_count_;
+} Config;
 
 //------------------------------------------------------------------------------
 // Helper functions.
 
-static int CountOccurrences(const char* arglist[], int list_length,
-                            const char* arg) {
+static int CountOccurrences(const CommandLineArguments* const args,
+                            const char* const arg) {
   int i;
   int num_occurences = 0;
 
-  for (i = 0; i < list_length; ++i) {
-    if (!strcmp(arglist[i], arg)) {
+  for (i = 0; i < args->argc_; ++i) {
+    if (!strcmp(args->argv_[i], arg)) {
       ++num_occurences;
     }
   }
@@ -301,6 +300,7 @@ static void PrintHelp(void) {
   printf("       webpmux -info INPUT\n");
   printf("       webpmux [-h|-help]\n");
   printf("       webpmux -version\n");
+  printf("       webpmux argument_file_name\n");
 
   printf("\n");
   printf("GET_OPTIONS:\n");
@@ -369,6 +369,10 @@ static void PrintHelp(void) {
 
   printf("\nNote: The nature of EXIF, XMP and ICC data is not checked");
   printf(" and is assumed to be\nvalid.\n");
+  printf("\nNote: if a single file name is passed as the argument, the "
+         "arguments will be\n");
+  printf("tokenized from this file. The file name must not start with "
+         "the character '-'.\n");
 }
 
 static void WarnAboutOddOffset(const WebPMuxFrameInfo* const info) {
@@ -379,22 +383,12 @@ static void WarnAboutOddOffset(const WebPMuxFrameInfo* const info) {
   }
 }
 
-static int ReadFileToWebPData(const char* const filename,
-                              WebPData* const webp_data) {
-  const uint8_t* data;
-  size_t size;
-  if (!ImgIoUtilReadFile(filename, &data, &size)) return 0;
-  webp_data->bytes = data;
-  webp_data->size = size;
-  return 1;
-}
-
 static int CreateMux(const char* const filename, WebPMux** mux) {
   WebPData bitstream;
   assert(mux != NULL);
-  if (!ReadFileToWebPData(filename, &bitstream)) return 0;
+  if (!ExUtilReadFileToWebPData(filename, &bitstream)) return 0;
   *mux = WebPMuxCreate(&bitstream, 1);
-  free((void*)bitstream.bytes);
+  WebPDataClear(&bitstream);
   if (*mux != NULL) return 1;
   fprintf(stderr, "Failed to create mux object from file %s.\n", filename);
   return 0;
@@ -517,9 +511,10 @@ static int ParseBgcolorArgs(const char* args, uint32_t* const bgcolor) {
 //------------------------------------------------------------------------------
 // Clean-up.
 
-static void DeleteConfig(WebPMuxConfig* config) {
+static void DeleteConfig(Config* const config) {
   if (config != NULL) {
-    free(config->feature_.args_);
+    free(config->args_);
+    ExUtilDeleteCommandLineArguments(&config->cmd_args_);
     memset(config, 0, sizeof(*config));
   }
 }
@@ -531,7 +526,7 @@ static void DeleteConfig(WebPMuxConfig* config) {
 // Returns 1 on valid, 0 otherwise.
 // Also fills up num_feature_args to be number of feature arguments given.
 // (e.g. if there are 4 '-frame's and 1 '-loop', then num_feature_args = 5).
-static int ValidateCommandLine(int argc, const char* argv[],
+static int ValidateCommandLine(const CommandLineArguments* const cmd_args,
                                int* num_feature_args) {
   int num_frame_args;
   int num_loop_args;
@@ -543,27 +538,27 @@ static int ValidateCommandLine(int argc, const char* argv[],
   *num_feature_args = 0;
 
   // Simple checks.
-  if (CountOccurrences(argv, argc, "-get") > 1) {
+  if (CountOccurrences(cmd_args, "-get") > 1) {
     ERROR_GOTO1("ERROR: Multiple '-get' arguments specified.\n", ErrValidate);
   }
-  if (CountOccurrences(argv, argc, "-set") > 1) {
+  if (CountOccurrences(cmd_args, "-set") > 1) {
     ERROR_GOTO1("ERROR: Multiple '-set' arguments specified.\n", ErrValidate);
   }
-  if (CountOccurrences(argv, argc, "-strip") > 1) {
+  if (CountOccurrences(cmd_args, "-strip") > 1) {
     ERROR_GOTO1("ERROR: Multiple '-strip' arguments specified.\n", ErrValidate);
   }
-  if (CountOccurrences(argv, argc, "-info") > 1) {
+  if (CountOccurrences(cmd_args, "-info") > 1) {
     ERROR_GOTO1("ERROR: Multiple '-info' arguments specified.\n", ErrValidate);
   }
-  if (CountOccurrences(argv, argc, "-o") > 1) {
+  if (CountOccurrences(cmd_args, "-o") > 1) {
     ERROR_GOTO1("ERROR: Multiple output files specified.\n", ErrValidate);
   }
 
   // Compound checks.
-  num_frame_args = CountOccurrences(argv, argc, "-frame");
-  num_loop_args = CountOccurrences(argv, argc, "-loop");
-  num_bgcolor_args = CountOccurrences(argv, argc, "-bgcolor");
-  num_durations_args = CountOccurrences(argv, argc, "-duration");
+  num_frame_args = CountOccurrences(cmd_args, "-frame");
+  num_loop_args = CountOccurrences(cmd_args, "-loop");
+  num_bgcolor_args = CountOccurrences(cmd_args, "-bgcolor");
+  num_durations_args = CountOccurrences(cmd_args, "-duration");
 
   if (num_loop_args > 1) {
     ERROR_GOTO1("ERROR: Multiple loop counts specified.\n", ErrValidate);
@@ -598,31 +593,35 @@ static int ValidateCommandLine(int argc, const char* argv[],
 
 #define ACTION_IS_NIL (config->action_type_ == NIL_ACTION)
 
-#define FEATURETYPE_IS_NIL (feature->type_ == NIL_FEATURE)
+#define FEATURETYPE_IS_NIL (config->type_ == NIL_FEATURE)
 
-#define CHECK_NUM_ARGS_LESS(NUM, LABEL)                                  \
+#define CHECK_NUM_ARGS_AT_LEAST(NUM, LABEL)                              \
   if (argc < i + (NUM)) {                                                \
     fprintf(stderr, "ERROR: Too few arguments for '%s'.\n", argv[i]);    \
     goto LABEL;                                                          \
   }
 
-#define CHECK_NUM_ARGS_NOT_EQUAL(NUM, LABEL)                             \
-  if (argc != i + (NUM)) {                                               \
+#define CHECK_NUM_ARGS_AT_MOST(NUM, LABEL)                               \
+  if (argc > i + (NUM)) {                                                \
     fprintf(stderr, "ERROR: Too many arguments for '%s'.\n", argv[i]);   \
     goto LABEL;                                                          \
   }
 
+#define CHECK_NUM_ARGS_EXACTLY(NUM, LABEL)                               \
+  CHECK_NUM_ARGS_AT_LEAST(NUM, LABEL);                                   \
+  CHECK_NUM_ARGS_AT_MOST(NUM, LABEL);
+
 // Parses command-line arguments to fill up config object. Also performs some
 // semantic checks.
-static int ParseCommandLine(int argc, const char* argv[],
-                            WebPMuxConfig* config) {
+static int ParseCommandLine(Config* config) {
   int i = 0;
   int feature_arg_index = 0;
   int ok = 1;
+  int argc = config->cmd_args_.argc_;
+  const char* const* argv = config->cmd_args_.argv_;
 
   while (i < argc) {
-    Feature* const feature = &config->feature_;
-    FeatureArg* const arg = &feature->args_[feature_arg_index];
+    FeatureArg* const arg = &config->args_[feature_arg_index];
     if (argv[i][0] == '-') {  // One of the action types or output.
       if (!strcmp(argv[i], "-set")) {
         if (ACTION_IS_NIL) {
@@ -632,14 +631,14 @@ static int ParseCommandLine(int argc, const char* argv[],
         }
         ++i;
       } else if (!strcmp(argv[i], "-duration")) {
-        CHECK_NUM_ARGS_LESS(2, ErrParse);
+        CHECK_NUM_ARGS_AT_LEAST(2, ErrParse);
         if (ACTION_IS_NIL || config->action_type_ == ACTION_DURATION) {
           config->action_type_ = ACTION_DURATION;
         } else {
           ERROR_GOTO1("ERROR: Multiple actions specified.\n", ErrParse);
         }
-        if (FEATURETYPE_IS_NIL || feature->type_ == FEATURE_DURATION) {
-          feature->type_ = FEATURE_DURATION;
+        if (FEATURETYPE_IS_NIL || config->type_ == FEATURE_DURATION) {
+          config->type_ = FEATURE_DURATION;
         } else {
           ERROR_GOTO1("ERROR: Multiple features specified.\n", ErrParse);
         }
@@ -656,20 +655,20 @@ static int ParseCommandLine(int argc, const char* argv[],
       } else if (!strcmp(argv[i], "-strip")) {
         if (ACTION_IS_NIL) {
           config->action_type_ = ACTION_STRIP;
-          feature->arg_count_ = 0;
+          config->arg_count_ = 0;
         } else {
           ERROR_GOTO1("ERROR: Multiple actions specified.\n", ErrParse);
         }
         ++i;
       } else if (!strcmp(argv[i], "-frame")) {
-        CHECK_NUM_ARGS_LESS(3, ErrParse);
+        CHECK_NUM_ARGS_AT_LEAST(3, ErrParse);
         if (ACTION_IS_NIL || config->action_type_ == ACTION_SET) {
           config->action_type_ = ACTION_SET;
         } else {
           ERROR_GOTO1("ERROR: Multiple actions specified.\n", ErrParse);
         }
-        if (FEATURETYPE_IS_NIL || feature->type_ == FEATURE_ANMF) {
-          feature->type_ = FEATURE_ANMF;
+        if (FEATURETYPE_IS_NIL || config->type_ == FEATURE_ANMF) {
+          config->type_ = FEATURE_ANMF;
         } else {
           ERROR_GOTO1("ERROR: Multiple features specified.\n", ErrParse);
         }
@@ -679,14 +678,14 @@ static int ParseCommandLine(int argc, const char* argv[],
         ++feature_arg_index;
         i += 3;
       } else if (!strcmp(argv[i], "-loop") || !strcmp(argv[i], "-bgcolor")) {
-        CHECK_NUM_ARGS_LESS(2, ErrParse);
+        CHECK_NUM_ARGS_AT_LEAST(2, ErrParse);
         if (ACTION_IS_NIL || config->action_type_ == ACTION_SET) {
           config->action_type_ = ACTION_SET;
         } else {
           ERROR_GOTO1("ERROR: Multiple actions specified.\n", ErrParse);
         }
-        if (FEATURETYPE_IS_NIL || feature->type_ == FEATURE_ANMF) {
-          feature->type_ = FEATURE_ANMF;
+        if (FEATURETYPE_IS_NIL || config->type_ == FEATURE_ANMF) {
+          config->type_ = FEATURE_ANMF;
         } else {
           ERROR_GOTO1("ERROR: Multiple features specified.\n", ErrParse);
         }
@@ -696,16 +695,16 @@ static int ParseCommandLine(int argc, const char* argv[],
         ++feature_arg_index;
         i += 2;
       } else if (!strcmp(argv[i], "-o")) {
-        CHECK_NUM_ARGS_LESS(2, ErrParse);
+        CHECK_NUM_ARGS_AT_LEAST(2, ErrParse);
         config->output_ = argv[i + 1];
         i += 2;
       } else if (!strcmp(argv[i], "-info")) {
-        CHECK_NUM_ARGS_NOT_EQUAL(2, ErrParse);
+        CHECK_NUM_ARGS_EXACTLY(2, ErrParse);
         if (config->action_type_ != NIL_ACTION) {
           ERROR_GOTO1("ERROR: Multiple actions specified.\n", ErrParse);
         } else {
           config->action_type_ = ACTION_INFO;
-          feature->arg_count_ = 0;
+          config->arg_count_ = 0;
           config->input_ = argv[i + 1];
         }
         i += 2;
@@ -741,13 +740,13 @@ static int ParseCommandLine(int argc, const char* argv[],
       if (!strcmp(argv[i], "icc") || !strcmp(argv[i], "exif") ||
           !strcmp(argv[i], "xmp")) {
         if (FEATURETYPE_IS_NIL) {
-          feature->type_ = (!strcmp(argv[i], "icc")) ? FEATURE_ICCP :
+          config->type_ = (!strcmp(argv[i], "icc")) ? FEATURE_ICCP :
               (!strcmp(argv[i], "exif")) ? FEATURE_EXIF : FEATURE_XMP;
         } else {
           ERROR_GOTO1("ERROR: Multiple features specified.\n", ErrParse);
         }
         if (config->action_type_ == ACTION_SET) {
-          CHECK_NUM_ARGS_LESS(2, ErrParse);
+          CHECK_NUM_ARGS_AT_LEAST(2, ErrParse);
           arg->filename_ = argv[i + 1];
           ++feature_arg_index;
           i += 2;
@@ -756,8 +755,8 @@ static int ParseCommandLine(int argc, const char* argv[],
         }
       } else if (!strcmp(argv[i], "frame") &&
                  (config->action_type_ == ACTION_GET)) {
-        CHECK_NUM_ARGS_LESS(2, ErrParse);
-        feature->type_ = FEATURE_ANMF;
+        CHECK_NUM_ARGS_AT_LEAST(2, ErrParse);
+        config->type_ = FEATURE_ANMF;
         arg->params_ = argv[i + 1];
         ++feature_arg_index;
         i += 2;
@@ -777,9 +776,8 @@ static int ParseCommandLine(int argc, const char* argv[],
 }
 
 // Additional checks after config is filled.
-static int ValidateConfig(WebPMuxConfig* config) {
+static int ValidateConfig(Config* const config) {
   int ok = 1;
-  Feature* const feature = &config->feature_;
 
   // Action.
   if (ACTION_IS_NIL) {
@@ -795,7 +793,7 @@ static int ValidateConfig(WebPMuxConfig* config) {
   if (config->input_ == NULL) {
     if (config->action_type_ != ACTION_SET) {
       ERROR_GOTO1("ERROR: No input file specified.\n", ErrValidate2);
-    } else if (feature->type_ != FEATURE_ANMF) {
+    } else if (config->type_ != FEATURE_ANMF) {
       ERROR_GOTO1("ERROR: No input file specified.\n", ErrValidate2);
     }
   }
@@ -811,27 +809,28 @@ static int ValidateConfig(WebPMuxConfig* config) {
 
 // Create config object from command-line arguments.
 static int InitializeConfig(int argc, const char* argv[],
-                            WebPMuxConfig* config) {
+                            Config* const config) {
   int num_feature_args = 0;
-  int ok = 1;
+  int ok;
 
-  assert(config != NULL);
   memset(config, 0, sizeof(*config));
 
+  ok = ExUtilInitCommandLineArguments(argc, argv, &config->cmd_args_);
+  if (!ok) return 0;
+
   // Validate command-line arguments.
-  if (!ValidateCommandLine(argc, argv, &num_feature_args)) {
+  if (!ValidateCommandLine(&config->cmd_args_, &num_feature_args)) {
     ERROR_GOTO1("Exiting due to command-line parsing error.\n", Err1);
   }
 
-  config->feature_.arg_count_ = num_feature_args;
-  config->feature_.args_ =
-      (FeatureArg*)calloc(num_feature_args, sizeof(*config->feature_.args_));
-  if (config->feature_.args_ == NULL) {
+  config->arg_count_ = num_feature_args;
+  config->args_ = (FeatureArg*)calloc(num_feature_args, sizeof(*config->args_));
+  if (config->args_ == NULL) {
     ERROR_GOTO1("ERROR: Memory allocation error.\n", Err1);
   }
 
   // Parse command-line.
-  if (!ParseCommandLine(argc, argv, config) || !ValidateConfig(config)) {
+  if (!ParseCommandLine(config) || !ValidateConfig(config)) {
     ERROR_GOTO1("Exiting due to command-line parsing error.\n", Err1);
   }
 
@@ -841,13 +840,14 @@ static int InitializeConfig(int argc, const char* argv[],
 
 #undef ACTION_IS_NIL
 #undef FEATURETYPE_IS_NIL
-#undef CHECK_NUM_ARGS_LESS
-#undef CHECK_NUM_ARGS_MORE
+#undef CHECK_NUM_ARGS_AT_LEAST
+#undef CHECK_NUM_ARGS_AT_MOST
+#undef CHECK_NUM_ARGS_EXACTLY
 
 //------------------------------------------------------------------------------
 // Processing.
 
-static int GetFrame(const WebPMux* mux, const WebPMuxConfig* config) {
+static int GetFrame(const WebPMux* mux, const Config* config) {
   WebPMuxError err = WEBP_MUX_OK;
   WebPMux* mux_single = NULL;
   int num = 0;
@@ -857,7 +857,7 @@ static int GetFrame(const WebPMux* mux, const WebPMuxConfig* config) {
   WebPMuxFrameInfo info;
   WebPDataInit(&info.bitstream);
 
-  num = ExUtilGetInt(config->feature_.args_[0].params_, 10, &parse_error);
+  num = ExUtilGetInt(config->args_[0].params_, 10, &parse_error);
   if (num < 0) {
     ERROR_GOTO1("ERROR: Frame/Fragment index must be non-negative.\n", ErrGet);
   }
@@ -891,18 +891,17 @@ static int GetFrame(const WebPMux* mux, const WebPMuxConfig* config) {
 }
 
 // Read and process config.
-static int Process(const WebPMuxConfig* config) {
+static int Process(const Config* config) {
   WebPMux* mux = NULL;
   WebPData chunk;
   WebPMuxError err = WEBP_MUX_OK;
   int ok = 1;
-  const Feature* const feature = &config->feature_;
 
   switch (config->action_type_) {
     case ACTION_GET: {
       ok = CreateMux(config->input_, &mux);
       if (!ok) goto Err2;
-      switch (feature->type_) {
+      switch (config->type_) {
         case FEATURE_ANMF:
           ok = GetFrame(mux, config);
           break;
@@ -910,10 +909,10 @@ static int Process(const WebPMuxConfig* config) {
         case FEATURE_ICCP:
         case FEATURE_EXIF:
         case FEATURE_XMP:
-          err = WebPMuxGetChunk(mux, kFourccList[feature->type_], &chunk);
+          err = WebPMuxGetChunk(mux, kFourccList[config->type_], &chunk);
           if (err != WEBP_MUX_OK) {
             ERROR_GOTO3("ERROR (%s): Could not get the %s.\n",
-                        ErrorString(err), kDescriptions[feature->type_], Err2);
+                        ErrorString(err), kDescriptions[config->type_], Err2);
           }
           ok = WriteData(config->output_, &chunk);
           break;
@@ -925,7 +924,7 @@ static int Process(const WebPMuxConfig* config) {
       break;
     }
     case ACTION_SET: {
-      switch (feature->type_) {
+      switch (config->type_) {
         case FEATURE_ANMF: {
           int i;
           WebPMuxAnimParams params = { 0xFFFFFFFF, 0 };
@@ -934,11 +933,11 @@ static int Process(const WebPMuxConfig* config) {
             ERROR_GOTO2("ERROR (%s): Could not allocate a mux object.\n",
                         ErrorString(WEBP_MUX_MEMORY_ERROR), Err2);
           }
-          for (i = 0; i < feature->arg_count_; ++i) {
-            switch (feature->args_[i].subtype_) {
+          for (i = 0; i < config->arg_count_; ++i) {
+            switch (config->args_[i].subtype_) {
               case SUBTYPE_BGCOLOR: {
                 uint32_t bgcolor;
-                ok = ParseBgcolorArgs(feature->args_[i].params_, &bgcolor);
+                ok = ParseBgcolorArgs(config->args_[i].params_, &bgcolor);
                 if (!ok) {
                   ERROR_GOTO1("ERROR: Could not parse the background color \n",
                               Err2);
@@ -949,7 +948,7 @@ static int Process(const WebPMuxConfig* config) {
               case SUBTYPE_LOOP: {
                 int parse_error = 0;
                 const int loop_count =
-                    ExUtilGetInt(feature->args_[i].params_, 10, &parse_error);
+                    ExUtilGetInt(config->args_[i].params_, 10, &parse_error);
                 if (loop_count < 0 || loop_count > 65535) {
                   // Note: This is only a 'necessary' condition for loop_count
                   // to be valid. The 'sufficient' conditioned in checked in
@@ -965,10 +964,10 @@ static int Process(const WebPMuxConfig* config) {
               case SUBTYPE_ANMF: {
                 WebPMuxFrameInfo frame;
                 frame.id = WEBP_CHUNK_ANMF;
-                ok = ReadFileToWebPData(feature->args_[i].filename_,
-                                        &frame.bitstream);
+                ok = ExUtilReadFileToWebPData(config->args_[i].filename_,
+                                              &frame.bitstream);
                 if (!ok) goto Err2;
-                ok = ParseFrameArgs(feature->args_[i].params_, &frame);
+                ok = ParseFrameArgs(config->args_[i].params_, &frame);
                 if (!ok) {
                   WebPDataClear(&frame.bitstream);
                   ERROR_GOTO1("ERROR: Could not parse frame properties.\n",
@@ -1001,13 +1000,13 @@ static int Process(const WebPMuxConfig* config) {
         case FEATURE_XMP: {
           ok = CreateMux(config->input_, &mux);
           if (!ok) goto Err2;
-          ok = ReadFileToWebPData(feature->args_[0].filename_, &chunk);
+          ok = ExUtilReadFileToWebPData(config->args_[0].filename_, &chunk);
           if (!ok) goto Err2;
-          err = WebPMuxSetChunk(mux, kFourccList[feature->type_], &chunk, 1);
+          err = WebPMuxSetChunk(mux, kFourccList[config->type_], &chunk, 1);
           free((void*)chunk.bytes);
           if (err != WEBP_MUX_OK) {
             ERROR_GOTO3("ERROR (%s): Could not set the %s.\n",
-                        ErrorString(err), kDescriptions[feature->type_], Err2);
+                        ErrorString(err), kDescriptions[config->type_], Err2);
           }
           break;
         }
@@ -1043,11 +1042,11 @@ static int Process(const WebPMuxConfig* config) {
         for (i = 0; i < num_frames; ++i) durations[i] = -1;
 
         // Parse intervals to process.
-        for (i = 0; i < feature->arg_count_; ++i) {
+        for (i = 0; i < config->arg_count_; ++i) {
           int k;
           int args[3];
           int duration, start, end;
-          const int nb_args = ExUtilGetInts(feature->args_[i].params_,
+          const int nb_args = ExUtilGetInts(config->args_[i].params_,
                                             10, 3, args);
           ok = (nb_args >= 1);
           if (!ok) goto Err3;
@@ -1105,12 +1104,12 @@ static int Process(const WebPMuxConfig* config) {
     case ACTION_STRIP: {
       ok = CreateMux(config->input_, &mux);
       if (!ok) goto Err2;
-      if (feature->type_ == FEATURE_ICCP || feature->type_ == FEATURE_EXIF ||
-          feature->type_ == FEATURE_XMP) {
-        err = WebPMuxDeleteChunk(mux, kFourccList[feature->type_]);
+      if (config->type_ == FEATURE_ICCP || config->type_ == FEATURE_EXIF ||
+          config->type_ == FEATURE_XMP) {
+        err = WebPMuxDeleteChunk(mux, kFourccList[config->type_]);
         if (err != WEBP_MUX_OK) {
           ERROR_GOTO3("ERROR (%s): Could not strip the %s.\n",
-                      ErrorString(err), kDescriptions[feature->type_], Err2);
+                      ErrorString(err), kDescriptions[config->type_], Err2);
         }
       } else {
         ERROR_GOTO1("ERROR: Invalid feature for action 'strip'.\n", Err2);
@@ -1140,7 +1139,7 @@ static int Process(const WebPMuxConfig* config) {
 // Main.
 
 int main(int argc, const char* argv[]) {
-  WebPMuxConfig config;
+  Config config;
   int ok = InitializeConfig(argc - 1, argv + 1, &config);
   if (ok) {
     ok = Process(&config);

extras/Makefile.am

@@ -13,7 +13,10 @@ libwebpextras_la_LDFLAGS = -lm
 libwebpextras_la_LIBADD = ../src/libwebp.la
 
 noinst_PROGRAMS =
-noinst_PROGRAMS += get_disto webp_quality
+noinst_PROGRAMS += webp_quality
+if BUILD_DEMUX
+  noinst_PROGRAMS += get_disto
+endif
 if BUILD_VWEBP_SDL
   noinst_PROGRAMS += vwebp_sdl
 endif
@@ -27,7 +30,7 @@ get_disto_LDADD += ../src/libwebp.la
 get_disto_LDADD += $(PNG_LIBS) $(JPEG_LIBS) $(TIFF_LIBS)
 
 webp_quality_SOURCES  = webp_quality.c
-webp_quality_CPPFLAGS = $(AM_CPPFLAGS) $(USE_EXPERIMENTAL_CODE)
+webp_quality_CPPFLAGS = $(AM_CPPFLAGS)
 webp_quality_LDADD =
 webp_quality_LDADD += ../imageio/libimageio_util.la
 webp_quality_LDADD += libwebpextras.la

extras/extras.c

@@ -16,8 +16,8 @@
 #include <assert.h>
 #include <string.h>
 
-#define XTRA_MAJ_VERSION 0
-#define XTRA_MIN_VERSION 1
+#define XTRA_MAJ_VERSION 1
+#define XTRA_MIN_VERSION 0
 #define XTRA_REV_VERSION 1
 
 //------------------------------------------------------------------------------
@@ -48,13 +48,14 @@ int WebPImportGray(const uint8_t* gray_data, WebPPicture* pic) {
 
 int WebPImportRGB565(const uint8_t* rgb565, WebPPicture* pic) {
   int x, y;
+  uint32_t* dst;
   if (pic == NULL || rgb565 == NULL) return 0;
   pic->colorspace = WEBP_YUV420;
   pic->use_argb = 1;
   if (!WebPPictureAlloc(pic)) return 0;
+  dst = pic->argb;
   for (y = 0; y < pic->height; ++y) {
     const int width = pic->width;
-    uint32_t* dst = pic->argb + y * pic->argb_stride;
     for (x = 0; x < width; ++x) {
 #ifdef WEBP_SWAP_16BIT_CSP
       const uint32_t rg = rgb565[2 * x + 1];
@@ -70,22 +71,24 @@ int WebPImportRGB565(const uint8_t* rgb565, WebPPicture* pic) {
       r = r | (r >> 5);
       g = g | (g >> 6);
       b = b | (b >> 5);
-      dst[x] = (r << 16) | (g << 8) | b;
+      dst[x] = (0xffu << 24) | (r << 16) | (g << 8) | b;
     }
     rgb565 += 2 * width;
+    dst += pic->argb_stride;
   }
   return 1;
 }
 
 int WebPImportRGB4444(const uint8_t* rgb4444, WebPPicture* pic) {
   int x, y;
+  uint32_t* dst;
   if (pic == NULL || rgb4444 == NULL) return 0;
   pic->colorspace = WEBP_YUV420;
   pic->use_argb = 1;
   if (!WebPPictureAlloc(pic)) return 0;
+  dst = pic->argb;
   for (y = 0; y < pic->height; ++y) {
     const int width = pic->width;
-    uint32_t* dst = pic->argb + y * pic->argb_stride;
     for (x = 0; x < width; ++x) {
 #ifdef WEBP_SWAP_16BIT_CSP
       const uint32_t rg = rgb4444[2 * x + 1];
@@ -106,6 +109,7 @@ int WebPImportRGB4444(const uint8_t* rgb4444, WebPPicture* pic) {
       dst[x] = (a << 24) | (r << 16) | (g << 8) | b;
     }
     rgb4444 += 2 * width;
+    dst += pic->argb_stride;
   }
   return 1;
 }

extras/extras.h

@@ -67,4 +67,4 @@ WEBP_EXTERN int VP8EstimateQuality(const uint8_t* const data, size_t size);
 }    // extern "C"
 #endif
 
-#endif  /* WEBP_EXTRAS_EXTRAS_H_ */
+#endif  // WEBP_EXTRAS_EXTRAS_H_

extras/get_disto.c

@@ -290,9 +290,10 @@ int main(int argc, const char *argv[]) {
     fprintf(stderr, "Error while computing the distortion.\n");
     goto End;
   }
-  printf("%u %.2f    %.2f %.2f %.2f %.2f\n",
+  printf("%u %.2f    %.2f %.2f %.2f %.2f [ %.2f bpp ]\n",
          (unsigned int)size1,
-         disto[4], disto[0], disto[1], disto[2], disto[3]);
+         disto[4], disto[0], disto[1], disto[2], disto[3],
+         8.f * size1 / pic1.width / pic1.height);
 
   if (output != NULL) {
     uint8_t* data = NULL;

extras/quality_estimate.c

@@ -73,7 +73,7 @@ int VP8EstimateQuality(const uint8_t* const data, size_t size) {
   pos += 4;
   bit_pos = pos * 8;
 
-  GET_BIT(2);  // color_space + clamp type
+  GET_BIT(2);  // colorspace + clamp type
 
   // Segment header
   if (GET_BIT(1)) {       // use_segment_

extras/webp_to_sdl.c

@@ -12,7 +12,7 @@
 // Author: James Zern (jzern@google.com)
 
 #ifdef HAVE_CONFIG_H
-#include "webp/config.h"
+#include "src/webp/config.h"
 #endif
 
 #if defined(WEBP_HAVE_SDL)
@@ -20,7 +20,7 @@
 #include "webp_to_sdl.h"
 
 #include <stdio.h>
-#include "webp/decode.h"
+#include "src/webp/decode.h"
 
 #if defined(WEBP_HAVE_JUST_SDL_H)
 #include <SDL.h>
@@ -40,7 +40,7 @@ int WebpToSDL(const char* data, unsigned int data_size) {
 
   if (!WebPInitDecoderConfig(&config)) {
     fprintf(stderr, "Library version mismatch!\n");
-    return 1;
+    return 0;
   }
 
   if (!init_ok) {

imageio/Makefile.am

@@ -1,7 +1,9 @@
 AM_CPPFLAGS += -I$(top_builddir)/src -I$(top_srcdir)/src
 noinst_LTLIBRARIES =
 noinst_LTLIBRARIES += libimageio_util.la
-noinst_LTLIBRARIES += libimagedec.la
+if BUILD_DEMUX
+  noinst_LTLIBRARIES += libimagedec.la
+endif
 noinst_LTLIBRARIES += libimageenc.la
 
 noinst_HEADERS =
@@ -21,9 +23,10 @@ libimagedec_la_SOURCES += tiffdec.c tiffdec.h
 libimagedec_la_SOURCES += webpdec.c webpdec.h
 libimagedec_la_SOURCES += wicdec.c wicdec.h
 libimagedec_la_CPPFLAGS = $(JPEG_INCLUDES) $(PNG_INCLUDES) $(TIFF_INCLUDES)
-libimagedec_la_CPPFLAGS += $(AM_CPPFLAGS) $(USE_EXPERIMENTAL_CODE)
+libimagedec_la_CPPFLAGS += $(AM_CPPFLAGS)
+libimagedec_la_LIBADD = ../src/demux/libwebpdemux.la
 
 libimageenc_la_SOURCES  =
 libimageenc_la_SOURCES += image_enc.c image_enc.h
 libimageenc_la_CPPFLAGS = $(JPEG_INCLUDES) $(PNG_INCLUDES) $(TIFF_INCLUDES)
-libimageenc_la_CPPFLAGS += $(AM_CPPFLAGS) $(USE_EXPERIMENTAL_CODE)
+libimageenc_la_CPPFLAGS += $(AM_CPPFLAGS)

imageio/image_enc.c

@@ -158,14 +158,8 @@ static void PNGAPI PNGErrorFunction(png_structp png, png_const_charp dummy) {
 }
 
 int WebPWritePNG(FILE* out_file, const WebPDecBuffer* const buffer) {
-  const uint32_t width = buffer->width;
-  const uint32_t height = buffer->height;
-  png_bytep row = buffer->u.RGBA.rgba;
-  const int stride = buffer->u.RGBA.stride;
-  const int has_alpha = WebPIsAlphaMode(buffer->colorspace);
   volatile png_structp png;
   volatile png_infop info;
-  png_uint_32 y;
 
   if (out_file == NULL || buffer == NULL) return 0;
 
@@ -184,14 +178,23 @@ int WebPWritePNG(FILE* out_file, const WebPDecBuffer* const buffer) {
     return 0;
   }
   png_init_io(png, out_file);
-  png_set_IHDR(png, info, width, height, 8,
-               has_alpha ? PNG_COLOR_TYPE_RGBA : PNG_COLOR_TYPE_RGB,
-               PNG_INTERLACE_NONE, PNG_COMPRESSION_TYPE_DEFAULT,
-               PNG_FILTER_TYPE_DEFAULT);
-  png_write_info(png, info);
-  for (y = 0; y < height; ++y) {
-    png_write_rows(png, &row, 1);
-    row += stride;
+  {
+    const uint32_t width = buffer->width;
+    const uint32_t height = buffer->height;
+    png_bytep row = buffer->u.RGBA.rgba;
+    const int stride = buffer->u.RGBA.stride;
+    const int has_alpha = WebPIsAlphaMode(buffer->colorspace);
+    uint32_t y;
+
+    png_set_IHDR(png, info, width, height, 8,
+                 has_alpha ? PNG_COLOR_TYPE_RGBA : PNG_COLOR_TYPE_RGB,
+                 PNG_INTERLACE_NONE, PNG_COMPRESSION_TYPE_DEFAULT,
+                 PNG_FILTER_TYPE_DEFAULT);
+    png_write_info(png, info);
+    for (y = 0; y < height; ++y) {
+      png_write_rows(png, &row, 1);
+      row += stride;
+    }
   }
   png_write_end(png, info);
   png_destroy_write_struct((png_structpp)&png, (png_infopp)&info);

imageio/imageio_util.c

@@ -47,7 +47,8 @@ int ImgIoUtilReadFromStdin(const uint8_t** data, size_t* data_size) {
   while (!feof(stdin)) {
     // We double the buffer size each time and read as much as possible.
     const size_t extra_size = (max_size == 0) ? kBlockSize : max_size;
-    void* const new_data = realloc(input, max_size + extra_size);
+    // we allocate one extra byte for the \0 terminator
+    void* const new_data = realloc(input, max_size + extra_size + 1);
     if (new_data == NULL) goto Error;
     input = (uint8_t*)new_data;
     max_size += extra_size;
@@ -55,6 +56,7 @@ int ImgIoUtilReadFromStdin(const uint8_t** data, size_t* data_size) {
     if (size < max_size) break;
   }
   if (ferror(stdin)) goto Error;
+  if (input != NULL) input[size] = '\0';  // convenient 0-terminator
   *data = input;
   *data_size = size;
   return 1;
@@ -68,7 +70,7 @@ int ImgIoUtilReadFromStdin(const uint8_t** data, size_t* data_size) {
 int ImgIoUtilReadFile(const char* const file_name,
                       const uint8_t** data, size_t* data_size) {
   int ok;
-  void* file_data;
+  uint8_t* file_data;
   size_t file_size;
   FILE* in;
   const int from_stdin = (file_name == NULL) || !strcmp(file_name, "-");
@@ -87,8 +89,14 @@ int ImgIoUtilReadFile(const char* const file_name,
   fseek(in, 0, SEEK_END);
   file_size = ftell(in);
   fseek(in, 0, SEEK_SET);
-  file_data = malloc(file_size);
-  if (file_data == NULL) return 0;
+  // we allocate one extra byte for the \0 terminator
+  file_data = (uint8_t*)malloc(file_size + 1);
+  if (file_data == NULL) {
+    fclose(in);
+    fprintf(stderr, "memory allocation failure when reading file %s\n",
+            file_name);
+    return 0;
+  }
   ok = (fread(file_data, file_size, 1, in) == 1);
   fclose(in);
 
@@ -98,11 +106,14 @@ int ImgIoUtilReadFile(const char* const file_name,
     free(file_data);
     return 0;
   }
-  *data = (uint8_t*)file_data;
+  file_data[file_size] = '\0';  // convenient 0-terminator
+  *data = file_data;
   *data_size = file_size;
   return 1;
 }
 
+// -----------------------------------------------------------------------------
+
 int ImgIoUtilWriteFile(const char* const file_name,
                        const uint8_t* data, size_t data_size) {
   int ok;

imageio/imageio_util.h

@@ -30,6 +30,9 @@ FILE* ImgIoUtilSetBinaryMode(FILE* file);
 // Allocates storage for entire file 'file_name' and returns contents and size
 // in 'data' and 'data_size'. Returns 1 on success, 0 otherwise. '*data' should
 // be deleted using free().
+// Note: for convenience, the data will be null-terminated with an extra byte
+// (not accounted for in *data_size), in case the file is text and intended
+// to be used as a C-string.
 // If 'file_name' is NULL or equal to "-", input is read from stdin by calling
 // the function ImgIoUtilReadFromStdin().
 int ImgIoUtilReadFile(const char* const file_name,

imageio/pngdec.c

@@ -18,6 +18,9 @@
 #include <stdio.h>
 
 #ifdef WEBP_HAVE_PNG
+#ifndef PNG_USER_MEM_SUPPORTED
+#define PNG_USER_MEM_SUPPORTED  // for png_create_read_struct_2
+#endif
 #include <png.h>
 #include <setjmp.h>   // note: this must be included *after* png.h
 #include <stdlib.h>
@@ -27,11 +30,33 @@
 #include "./imageio_util.h"
 #include "./metadata.h"
 
+#define LOCAL_PNG_VERSION ((PNG_LIBPNG_VER_MAJOR << 8) | PNG_LIBPNG_VER_MINOR)
+#define LOCAL_PNG_PREREQ(maj, min) \
+   (LOCAL_PNG_VERSION >= (((maj) << 8) | (min)))
+
 static void PNGAPI error_function(png_structp png, png_const_charp error) {
   if (error != NULL) fprintf(stderr, "libpng error: %s\n", error);
   longjmp(png_jmpbuf(png), 1);
 }
 
+#if LOCAL_PNG_PREREQ(1,4)
+typedef png_alloc_size_t LocalPngAllocSize;
+#else
+typedef png_size_t LocalPngAllocSize;
+#endif
+
+static png_voidp MallocFunc(png_structp png_ptr, LocalPngAllocSize size) {
+  (void)png_ptr;
+  if (size != (size_t)size) return NULL;
+  if (!ImgIoUtilCheckSizeArgumentsOverflow(size, 1)) return NULL;
+  return (png_voidp)malloc((size_t)size);
+}
+
+static void FreeFunc(png_structp png_ptr, png_voidp ptr) {
+  (void)png_ptr;
+  free(ptr);
+}
+
 // Converts the NULL terminated 'hexstring' which contains 2-byte character
 // representations of hex values to raw data.
 // 'hexstring' may contain values consisting of [A-F][a-f][0-9] in pairs,
@@ -171,11 +196,10 @@ static int ExtractMetadataFromPNG(png_structp png,
     {
       png_charp name;
       int comp_type;
-#if ((PNG_LIBPNG_VER_MAJOR << 8) | PNG_LIBPNG_VER_MINOR << 0) < \
-    ((1 << 8) | (5 << 0))
-      png_charp profile;
-#else  // >= libpng 1.5.0
+#if LOCAL_PNG_PREREQ(1,5)
       png_bytep profile;
+#else
+      png_charp profile;
 #endif
       png_uint_32 len;
 
@@ -185,7 +209,6 @@ static int ExtractMetadataFromPNG(png_structp png,
       }
     }
   }
-
   return 1;
 }
 
@@ -225,7 +248,8 @@ int ReadPNG(const uint8_t* const data, size_t data_size,
   context.data = data;
   context.data_size = data_size;
 
-  png = png_create_read_struct(PNG_LIBPNG_VER_STRING, 0, 0, 0);
+  png = png_create_read_struct_2(PNG_LIBPNG_VER_STRING, NULL, NULL, NULL,
+                                 NULL, MallocFunc, FreeFunc);
   if (png == NULL) goto End;
 
   png_set_error_fn(png, 0, error_function, NULL);
@@ -265,6 +289,16 @@ int ReadPNG(const uint8_t* const data, size_t data_size,
     has_alpha = !!(color_type & PNG_COLOR_MASK_ALPHA);
   }
 
+  // Apply gamma correction if needed.
+  {
+    double image_gamma = 1 / 2.2, screen_gamma = 2.2;
+    int srgb_intent;
+    if (png_get_sRGB(png, info, &srgb_intent) ||
+        png_get_gAMA(png, info, &image_gamma)) {
+      png_set_gamma(png, screen_gamma, image_gamma);
+    }
+  }
+
   if (!keep_alpha) {
     png_set_strip_alpha(png);
     has_alpha = 0;

imageio/pnmdec.c

@@ -117,7 +117,7 @@ static size_t ReadPAMFields(PNMInfo* const info, size_t off) {
     }
   }
   if (!(info->seen_flags & TUPLE_FLAG)) {
-    if (info->depth > 0 && info->depth <= 4) {
+    if (info->depth > 0 && info->depth <= 4 && info->depth != 2) {
       info->seen_flags |= TUPLE_FLAG;
       info->bytes_per_px = info->depth * (info->max_value > 255 ? 2 : 1);
     } else {
@@ -165,7 +165,7 @@ static size_t ReadHeader(PNMInfo* const info) {
   // perform some basic numerical validation
   if (info->width <= 0 || info->height <= 0 ||
       info->type <= 0 || info->type >= 9 ||
-      info->depth <= 0 || info->depth > 4 ||
+      info->depth <= 0 || info->depth == 2 || info->depth > 4 ||
       info->bytes_per_px < info->depth ||
       info->max_value <= 0 || info->max_value >= 65536) {
     return 0;

imageio/tiffdec.c

@@ -121,7 +121,6 @@ static tsize_t MyRead(thandle_t opaque, void* dst, tsize_t size) {
 // (we don't want to force a dependency to a libdspdec library).
 #define MFIX 24    // 24bit fixed-point arithmetic
 #define HALF ((1u << MFIX) >> 1)
-#define KINV_255 ((1u << MFIX) / 255u)
 
 static uint32_t Unmult(uint8_t x, uint32_t mult) {
   const uint32_t v = (x * mult + HALF) >> MFIX;
@@ -132,6 +131,9 @@ static WEBP_INLINE uint32_t GetScale(uint32_t a) {
   return (255u << MFIX) / a;
 }
 
+#undef MFIX
+#undef HALF
+
 static void MultARGBRow(uint8_t* ptr, int width) {
   int x;
   for (x = 0; x < width; ++x, ptr += 4) {

imageio/webpdec.c

@@ -15,10 +15,12 @@
 
 #include "./webpdec.h"
 
+#include <assert.h>
 #include <stdio.h>
 #include <stdlib.h>
 
 #include "webp/decode.h"
+#include "webp/demux.h"
 #include "webp/encode.h"
 #include "./imageio_util.h"
 #include "./metadata.h"
@@ -95,25 +97,47 @@ VP8StatusCode DecodeWebPIncremental(
       fprintf(stderr, "Failed during WebPINewDecoder().\n");
       return VP8_STATUS_OUT_OF_MEMORY;
     } else {
-#ifdef WEBP_EXPERIMENTAL_FEATURES
-      size_t size = 0;
-      const size_t incr = 2 + (data_size / 20);
-      while (size < data_size) {
-        size_t next_size = size + (rand() % incr);
-        if (next_size > data_size) next_size = data_size;
-        status = WebPIUpdate(idec, data, next_size);
-        if (status != VP8_STATUS_OK && status != VP8_STATUS_SUSPENDED) break;
-        size = next_size;
-      }
-#else
       status = WebPIUpdate(idec, data, data_size);
-#endif
       WebPIDelete(idec);
     }
   }
   return status;
 }
 
+// -----------------------------------------------------------------------------
+// Metadata
+
+static int ExtractMetadata(const uint8_t* const data, size_t data_size,
+                           Metadata* const metadata) {
+  WebPData webp_data = { data, data_size };
+  WebPDemuxer* const demux = WebPDemux(&webp_data);
+  WebPChunkIterator chunk_iter;
+  uint32_t flags;
+
+  if (demux == NULL) return 0;
+  assert(metadata != NULL);
+
+  flags = WebPDemuxGetI(demux, WEBP_FF_FORMAT_FLAGS);
+
+  if ((flags & ICCP_FLAG) && WebPDemuxGetChunk(demux, "ICCP", 1, &chunk_iter)) {
+    MetadataCopy((const char*)chunk_iter.chunk.bytes, chunk_iter.chunk.size,
+                 &metadata->iccp);
+    WebPDemuxReleaseChunkIterator(&chunk_iter);
+  }
+  if ((flags & EXIF_FLAG) && WebPDemuxGetChunk(demux, "EXIF", 1, &chunk_iter)) {
+    MetadataCopy((const char*)chunk_iter.chunk.bytes, chunk_iter.chunk.size,
+                 &metadata->exif);
+    WebPDemuxReleaseChunkIterator(&chunk_iter);
+  }
+  if ((flags & XMP_FLAG) && WebPDemuxGetChunk(demux, "XMP ", 1, &chunk_iter)) {
+    MetadataCopy((const char*)chunk_iter.chunk.bytes, chunk_iter.chunk.size,
+                 &metadata->xmp);
+    WebPDemuxReleaseChunkIterator(&chunk_iter);
+  }
+  WebPDemuxDelete(demux);
+  return 1;
+}
+
 // -----------------------------------------------------------------------------
 
 int ReadWebP(const uint8_t* const data, size_t data_size,
@@ -127,11 +151,6 @@ int ReadWebP(const uint8_t* const data, size_t data_size,
 
   if (data == NULL || data_size == 0 || pic == NULL) return 0;
 
-  // TODO(jzern): add Exif/XMP/ICC extraction.
-  if (metadata != NULL) {
-    fprintf(stderr, "Warning: metadata extraction from WebP is unsupported.\n");
-  }
-
   if (!WebPInitDecoderConfig(&config)) {
     fprintf(stderr, "Library version mismatch!\n");
     return 0;
@@ -193,7 +212,6 @@ int ReadWebP(const uint8_t* const data, size_t data_size,
 
     status = DecodeWebP(data, data_size, &config);
     ok = (status == VP8_STATUS_OK);
-    if (!ok) WebPPictureFree(pic);
     if (ok && !keep_alpha && pic->use_argb) {
       // Need to wipe out the alpha value, as requested.
       int x, y;
@@ -207,9 +225,18 @@ int ReadWebP(const uint8_t* const data, size_t data_size,
 
   if (status != VP8_STATUS_OK) {
     PrintWebPError("input data", status);
+    ok = 0;
   }
 
   WebPFreeDecBuffer(output_buffer);
+
+  if (ok && metadata != NULL) {
+    ok = ExtractMetadata(data, data_size, metadata);
+    if (!ok) {
+      PrintWebPError("metadata", VP8_STATUS_BITSTREAM_ERROR);
+    }
+  }
+  if (!ok) WebPPictureFree(pic);
   return ok;
 }
 

makefile.unix

@@ -25,6 +25,7 @@ ifeq ($(strip $(shell uname)), Darwin)
   # Failure observed with: gcc 4.2.1 and 4.0.1.
   EXTRA_FLAGS += -fno-common
   EXTRA_FLAGS += -DHAVE_GLUT_GLUT_H
+  EXTRA_FLAGS += -Wno-deprecated-declarations
   EXTRA_FLAGS += -I/opt/local/include
   EXTRA_LIBS  += -L/opt/local/lib
   GL_LIBS = -framework GLUT -framework OpenGL
@@ -63,9 +64,6 @@ endif
 # 'make -f makefile.unix EXTRA_FLAGS=-m32' to that effect.
 # EXTRA_FLAGS += -m32
 
-# Extra flags to enable experimental features and code
-# EXTRA_FLAGS += -DWEBP_EXPERIMENTAL_FEATURES
-
 # Extra flags to enable byte swap for 16 bit colorspaces.
 # EXTRA_FLAGS += -DWEBP_SWAP_16BIT_CSP=1
 
@@ -91,12 +89,6 @@ EXTRA_FLAGS += -DWEBP_HAVE_SSE41
 src/dsp/%_sse41.o: EXTRA_FLAGS += -msse4.1
 endif
 
-# AVX2-specific flags:
-ifeq ($(HAVE_AVX2), 1)
-EXTRA_FLAGS += -DWEBP_HAVE_AVX2
-src/dsp/%_avx2.o: EXTRA_FLAGS += -mavx2
-endif
-
 # NEON-specific flags:
 # EXTRA_FLAGS += -march=armv7-a -mfloat-abi=hard -mfpu=neon -mtune=cortex-a8
 # -> seems to make the overall lib slower: -fno-split-wide-types
@@ -114,7 +106,12 @@ endif
 AR = ar
 ARFLAGS = r
 CPPFLAGS = -I. -Isrc/ -Wall
-CFLAGS = -O3 -DNDEBUG $(EXTRA_FLAGS)
+ifeq ($(DEBUG), 1)
+  CFLAGS = -g
+else
+  CFLAGS = -O3 -DNDEBUG
+endif
+CFLAGS += $(EXTRA_FLAGS)
 CC = gcc
 INSTALL = install
 GROFF = /usr/bin/groff
@@ -176,11 +173,13 @@ DSP_DEC_OBJS = \
     src/dsp/upsampling_msa.o \
     src/dsp/upsampling_neon.o \
     src/dsp/upsampling_sse2.o \
+    src/dsp/upsampling_sse41.o \
     src/dsp/yuv.o \
     src/dsp/yuv_mips32.o \
     src/dsp/yuv_mips_dsp_r2.o \
     src/dsp/yuv_neon.o \
     src/dsp/yuv_sse2.o \
+    src/dsp/yuv_sse41.o \
 
 DSP_ENC_OBJS = \
     src/dsp/cost.o \
@@ -188,7 +187,6 @@ DSP_ENC_OBJS = \
     src/dsp/cost_mips_dsp_r2.o \
     src/dsp/cost_sse2.o \
     src/dsp/enc.o \
-    src/dsp/enc_avx2.o \
     src/dsp/enc_mips32.o \
     src/dsp/enc_mips_dsp_r2.o \
     src/dsp/enc_msa.o \
@@ -212,7 +210,6 @@ ENC_OBJS = \
     src/enc/backward_references_enc.o \
     src/enc/config_enc.o \
     src/enc/cost_enc.o \
-    src/enc/delta_palettization_enc.o \
     src/enc/filter_enc.o \
     src/enc/frame_enc.o \
     src/enc/histogram_enc.o \
@@ -310,7 +307,6 @@ HDRS = \
     src/dsp/yuv.h \
     src/enc/backward_references_enc.h \
     src/enc/cost_enc.h \
-    src/enc/delta_palettization_enc.h \
     src/enc/histogram_enc.h \
     src/enc/vp8i_enc.h \
     src/enc/vp8li_enc.h \
@@ -389,7 +385,7 @@ src/demux/libwebpdemux.a: $(LIBWEBPDEMUX_OBJS)
 	$(AR) $(ARFLAGS) $@ $^
 
 examples/anim_diff: examples/anim_diff.o $(ANIM_UTIL_OBJS) $(GIFDEC_OBJS)
-examples/anim_dump: examples/anim_dump.o $(ANIM_UTIL_OBJS)
+examples/anim_dump: examples/anim_dump.o $(ANIM_UTIL_OBJS) $(GIFDEC_OBJS)
 examples/cwebp: examples/cwebp.o
 examples/dwebp: examples/dwebp.o
 examples/gif2webp: examples/gif2webp.o $(GIFDEC_OBJS)
@@ -403,7 +399,7 @@ examples/anim_diff: src/demux/libwebpdemux.a examples/libexample_util.a
 examples/anim_diff: imageio/libimageio_util.a src/libwebp.a
 examples/anim_diff: EXTRA_LIBS += $(GIF_LIBS)
 examples/anim_diff: EXTRA_FLAGS += -DWEBP_HAVE_GIF
-examples/anim_dump: examples/libanim_util.a
+examples/anim_dump: examples/libanim_util.a examples/libgifdec.a
 examples/anim_dump: src/demux/libwebpdemux.a
 examples/anim_dump: examples/libexample_util.a
 examples/anim_dump: imageio/libimageio_util.a
@@ -412,11 +408,13 @@ examples/anim_dump: src/libwebp.a
 examples/anim_dump: EXTRA_LIBS += $(GIF_LIBS) $(DWEBP_LIBS)
 examples/cwebp: examples/libexample_util.a
 examples/cwebp: imageio/libimagedec.a
+examples/cwebp: src/demux/libwebpdemux.a
 examples/cwebp: imageio/libimageio_util.a
 examples/cwebp: src/libwebp.a
 examples/cwebp: EXTRA_LIBS += $(CWEBP_LIBS)
 examples/dwebp: examples/libexample_util.a
 examples/dwebp: imageio/libimagedec.a
+examples/dwebp: src/demux/libwebpdemux.a
 examples/dwebp: imageio/libimageenc.a
 examples/dwebp: imageio/libimageio_util.a
 examples/dwebp: src/libwebp.a
@@ -433,13 +431,17 @@ examples/webpmux: examples/libexample_util.a imageio/libimageio_util.a
 examples/webpmux: src/mux/libwebpmux.a src/libwebpdecoder.a
 examples/img2webp: examples/libexample_util.a imageio/libimageio_util.a
 examples/img2webp: imageio/libimagedec.a
+examples/img2webp: src/demux/libwebpdemux.a
 examples/img2webp: src/mux/libwebpmux.a src/libwebp.a
 examples/img2webp: EXTRA_LIBS += $(CWEBP_LIBS)
 examples/webpinfo: examples/libexample_util.a imageio/libimageio_util.a
 examples/webpinfo: src/libwebpdecoder.a
 
 extras/get_disto: extras/get_disto.o
-extras/get_disto: imageio/libimagedec.a imageio/libimageio_util.a src/libwebp.a
+extras/get_disto: imageio/libimagedec.a
+extras/get_disto: src/demux/libwebpdemux.a
+extras/get_disto: imageio/libimageio_util.a
+extras/get_disto: src/libwebp.a
 extras/get_disto: EXTRA_LIBS += $(CWEBP_LIBS)
 
 extras/webp_quality: extras/webp_quality.o
@@ -470,7 +472,7 @@ dist: all
 	for m in man/[cdv]webp.1 man/gif2webp.1 man/webpmux.1 \
                  man/img2webp.1 man/webpinfo.1; do \
 	  basenam=$$(basename $$m .1); \
-	  $(GROFF) -t -e -man -T utf8 $$m \
+	  $(GROFF) -t -e -man -T ascii $$m \
 	    | $(COL) -bx >$(DESTDIR)/doc/$${basenam}.txt; \
 	  $(GROFF) -t -e -man -T html $$m \
 	    | $(COL) -bx >$(DESTDIR)/doc/$${basenam}.html; \

man/Makefile.am

@@ -1,10 +1,13 @@
 man_MANS = cwebp.1 dwebp.1
-if WANT_MUX
+if BUILD_MUX
   man_MANS += webpmux.1
 endif
 if BUILD_GIF2WEBP
   man_MANS += gif2webp.1
 endif
+if BUILD_IMG2WEBP
+  man_MANS += img2webp.1
+endif
 if BUILD_VWEBP
   man_MANS += vwebp.1
 endif

man/gif2webp.1

@@ -1,5 +1,5 @@
 .\"                                      Hey, EMACS: -*- nroff -*-
-.TH GIF2WEBP 1 "September 20, 2017"
+.TH GIF2WEBP 1 "January 25, 2018"
 .SH NAME
 gif2webp \- Convert a GIF image to WebP
 .SH SYNOPSIS
@@ -20,6 +20,12 @@ Specify the name of the output WebP file. If omitted, \fBgif2webp\fP will
 perform conversion but only report statistics.
 Using "\-" as output name will direct output to 'stdout'.
 .TP
+.BI \-\- " string
+Explicitly specify the input file. This option is useful if the input
+file starts with an '\-' for instance. This option must appear \fBlast\fP.
+Any other options afterward will be ignored. If the input file is "\-",
+the data will be read from \fIstdin\fP instead of a file.
+.TP
 .B \-h, \-help
 Usage information.
 .TP
@@ -109,6 +115,7 @@ the range of 20 to 50.
 .TP
 .B \-mt
 Use multi-threading for encoding, if possible.
+.TP
 .B \-loop_compatibility
 If enabled, handle the loop information in a compatible fashion for Chrome
 version prior to M62 (inclusive) and Firefox.
@@ -136,6 +143,8 @@ gif2webp \-lossy \-m 3 picture.gif \-o picture_lossy.webp
 gif2webp \-lossy \-f 50 picture.gif \-o picture.webp
 .br
 gif2webp \-q 70 \-o picture.webp \-\- \-\-\-picture.gif
+.br
+cat picture.gif | gif2webp \-o \- \-\- \- > output.webp
 
 .SH AUTHORS
 \fBgif2webp\fP is a part of libwebp and was written by the WebP team.

man/img2webp.1

@@ -1,11 +1,13 @@
 .\"                                      Hey, EMACS: -*- nroff -*-
-.TH IMG2WEBP 1 "January 23, 2017"
+.TH IMG2WEBP 1 "April 3, 2018"
 .SH NAME
 img2webp \- create animated WebP file from a sequence of input images.
 .SH SYNOPSIS
 .B img2webp
 [file_level_options] [files] [per_frame_options...]
 .br
+.B img2webp argument_file_name
+.br
 .SH DESCRIPTION
 This manual page documents the
 .B img2webp
@@ -13,6 +15,9 @@ command.
 .PP
 \fBimg2webp\fP compresses a sequence of images using the animated WebP format.
 Input images can either be PNG, JPEG, TIFF or WebP.
+If a single file name (not starting with the character '\-') is supplied as
+the argument, the command line argument are actually tokenized from this file.
+This allows for easy scripting or using large number of arguments.
 .SH FILE-LEVEL OPTIONS
 The file-level options are applied at the beginning of the compression process,
 before the input frames are read.
@@ -40,14 +45,17 @@ lossy or lossless compression for each frame heuristically. This global
 option disables the local option \fB-lossy\fP and \fB-lossless\fP .
 .TP
 .BI \-loop " int
-Specifies the number of times the animation should loop. Using '0' means
-'loop indefinitely'.
+Specifies the number of times the animation should loop. Using '0'
+means 'loop indefinitely'.
 .TP
 .BI \-v
 Be more verbose.
 .TP
 .B \-h, \-help
 A short usage summary.
+.TP
+.B \-version
+Print the version numbers of the relevant libraries used.
 
 .SH PER-FRAME OPTIONS
 The per-frame options are applied for the images following as arguments in the

man/vwebp.1

@@ -1,5 +1,5 @@
 .\"                                      Hey, EMACS: -*- nroff -*-
-.TH VWEBP 1 "November 25, 2016"
+.TH VWEBP 1 "July 20, 2018"
 .SH NAME
 vwebp \- decompress a WebP file and display it in a window
 .SH SYNOPSIS
@@ -38,6 +38,10 @@ It helps by smoothing gradients and avoiding banding artifacts. Default: 50.
 By default, quantized transparency planes are dithered during decompression,
 to smooth the gradients. This flag will prevent this dithering.
 .TP
+.B \-usebgcolor
+Fill transparent areas with the bitstream's own background color instead of
+checkerboard only. Default is white for non-animated images.
+.TP
 .B \-mt
 Use multi-threading for decoding, if possible.
 .TP
@@ -56,6 +60,9 @@ the data will be read from \fIstdin\fP instead of a file.
 .B 'c'
 Toggle use of color profile.
 .TP
+.B 'b'
+Toggle display of background color.
+.TP
 .B 'i'
 Overlay file information.
 .TP

man/webpmux.1

@@ -1,5 +1,5 @@
 .\"                                      Hey, EMACS: -*- nroff -*-
-.TH WEBPMUX 1 "November 10, 2016"
+.TH WEBPMUX 1 "April 23, 2018"
 .SH NAME
 webpmux \- create animated WebP files from non\-animated WebP images, extract
 frames from animated WebP images, and manage XMP/EXIF metadata and ICC profile.
@@ -48,6 +48,8 @@ frames from animated WebP images, and manage XMP/EXIF metadata and ICC profile.
 .B webpmux [\-h|\-help]
 .br
 .B webpmux \-version
+.br
+.B webpmux argument_file_name
 .SH DESCRIPTION
 This manual page documents the
 .B webpmux
@@ -55,6 +57,9 @@ command.
 .PP
 \fBwebpmux\fP can be used to create/extract from animated WebP files, as well as
 to add/extract/strip XMP/EXIF metadata and ICC profile.
+If a single file name (not starting with the character '\-') is supplied as
+the argument, the command line arguments are actually tokenized from this file.
+This allows for easy scripting or using large number of arguments.
 .SH OPTIONS
 .SS GET_OPTIONS (\-get):
 .TP

src/Makefile.am

@@ -1,10 +1,10 @@
 # The mux and demux libraries depend on libwebp, thus the '.' to force
 # the build order so it's available to them.
 SUBDIRS = dec enc dsp utils .
-if WANT_MUX
+if BUILD_MUX
   SUBDIRS += mux
 endif
-if WANT_DEMUX
+if BUILD_DEMUX
   SUBDIRS += demux
 endif
 
@@ -36,7 +36,7 @@ libwebp_la_LIBADD += utils/libwebputils.la
 # other than the ones listed on the command line, i.e., after linking, it will
 # not have unresolved symbols. Some platforms (Windows among them) require all
 # symbols in shared libraries to be resolved at library creation.
-libwebp_la_LDFLAGS = -no-undefined -version-info 7:1:0
+libwebp_la_LDFLAGS = -no-undefined -version-info 7:3:0
 libwebpincludedir = $(includedir)/webp
 pkgconfig_DATA = libwebp.pc
 
@@ -48,7 +48,7 @@ if BUILD_LIBWEBPDECODER
   libwebpdecoder_la_LIBADD += dsp/libwebpdspdecode.la
   libwebpdecoder_la_LIBADD += utils/libwebputilsdecode.la
 
-  libwebpdecoder_la_LDFLAGS = -no-undefined -version-info 3:1:0
+  libwebpdecoder_la_LDFLAGS = -no-undefined -version-info 3:3:0
   pkgconfig_DATA += libwebpdecoder.pc
 endif
 

src/dec/Makefile.am

@@ -25,5 +25,5 @@ libwebpdecodeinclude_HEADERS += ../webp/types.h
 noinst_HEADERS =
 noinst_HEADERS += ../webp/format_constants.h
 
-libwebpdecode_la_CPPFLAGS = $(AM_CPPFLAGS) $(USE_EXPERIMENTAL_CODE)
+libwebpdecode_la_CPPFLAGS = $(AM_CPPFLAGS)
 libwebpdecodeincludedir = $(includedir)/webp

src/dec/alphai_dec.h

@@ -51,4 +51,4 @@ void WebPDeallocateAlphaMemory(VP8Decoder* const dec);
 }    // extern "C"
 #endif
 
-#endif  /* WEBP_DEC_ALPHAI_DEC_H_ */
+#endif  // WEBP_DEC_ALPHAI_DEC_H_

src/dec/common_dec.h

@@ -51,4 +51,4 @@ enum { MB_FEATURE_TREE_PROBS = 3,
        NUM_PROBAS = 11
      };
 
-#endif    // WEBP_DEC_COMMON_DEC_H_
+#endif  // WEBP_DEC_COMMON_DEC_H_

src/dec/frame_dec.c

@@ -338,7 +338,6 @@ void VP8InitDithering(const WebPDecoderOptions* const options,
       for (s = 0; s < NUM_MB_SEGMENTS; ++s) {
         VP8QuantMatrix* const dqm = &dec->dqm_[s];
         if (dqm->uv_quant_ < DITHER_AMP_TAB_SIZE) {
-          // TODO(skal): should we specially dither more for uv_quant_ < 0?
           const int idx = (dqm->uv_quant_ < 0) ? 0 : dqm->uv_quant_;
           dqm->dither_ = (f * kQuantToDitherAmp[idx]) >> 3;
         }
@@ -400,7 +399,9 @@ static void DitherRow(VP8Decoder* const dec) {
 #define MACROBLOCK_VPOS(mb_y)  ((mb_y) * 16)    // vertical position of a MB
 
 // Finalize and transmit a complete row. Return false in case of user-abort.
-static int FinishRow(VP8Decoder* const dec, VP8Io* const io) {
+static int FinishRow(void* arg1, void* arg2) {
+  VP8Decoder* const dec = (VP8Decoder*)arg1;
+  VP8Io* const io = (VP8Io*)arg2;
   int ok = 1;
   const VP8ThreadContext* const ctx = &dec->thread_ctx_;
   const int cache_id = ctx->id_;
@@ -448,10 +449,9 @@ static int FinishRow(VP8Decoder* const dec, VP8Io* const io) {
     if (y_end > io->crop_bottom) {
       y_end = io->crop_bottom;    // make sure we don't overflow on last row.
     }
+    // If dec->alpha_data_ is not NULL, we have some alpha plane present.
     io->a = NULL;
     if (dec->alpha_data_ != NULL && y_start < y_end) {
-      // TODO(skal): testing presence of alpha with dec->alpha_data_ is not a
-      // good idea.
       io->a = VP8DecompressAlphaRows(dec, io, y_start, y_end - y_start);
       if (io->a == NULL) {
         return VP8SetError(dec, VP8_STATUS_BITSTREAM_ERROR,
@@ -558,7 +558,6 @@ VP8StatusCode VP8EnterCritical(VP8Decoder* const dec, VP8Io* const io) {
   if (io->bypass_filtering) {
     dec->filter_type_ = 0;
   }
-  // TODO(skal): filter type / strength / sharpness forcing
 
   // Define the area where we can skip in-loop filtering, in case of cropping.
   //
@@ -569,8 +568,6 @@ VP8StatusCode VP8EnterCritical(VP8Decoder* const dec, VP8Io* const io) {
   // Means: there's a dependency chain that goes all the way up to the
   // top-left corner of the picture (MB #0). We must filter all the previous
   // macroblocks.
-  // TODO(skal): add an 'approximate_decoding' option, that won't produce
-  // a 1:1 bit-exactness for complex filtering?
   {
     const int extra_pixels = kFilterExtraRows[dec->filter_type_];
     if (dec->filter_type_ == 2) {
@@ -651,7 +648,7 @@ static int InitThreadContext(VP8Decoder* const dec) {
     }
     worker->data1 = dec;
     worker->data2 = (void*)&dec->thread_ctx_.io_;
-    worker->hook = (WebPWorkerHook)FinishRow;
+    worker->hook = FinishRow;
     dec->num_caches_ =
       (dec->filter_type_ > 0) ? MT_CACHE_LINES : MT_CACHE_LINES - 1;
   } else {
@@ -671,15 +668,9 @@ int VP8GetThreadMethod(const WebPDecoderOptions* const options,
   (void)height;
   assert(headers == NULL || !headers->is_lossless);
 #if defined(WEBP_USE_THREAD)
-  if (width < MIN_WIDTH_FOR_THREADS) return 0;
-  // TODO(skal): tune the heuristic further
-#if 0
-  if (height < 2 * width) return 2;
+  if (width >= MIN_WIDTH_FOR_THREADS) return 2;
 #endif
-  return 2;
-#else   // !WEBP_USE_THREAD
   return 0;
-#endif
 }
 
 #undef MT_CACHE_LINES

src/dec/idec_dec.c

@@ -140,10 +140,9 @@ static void DoRemap(WebPIDecoder* const idec, ptrdiff_t offset) {
       if (NeedCompressedAlpha(idec)) {
         ALPHDecoder* const alph_dec = dec->alph_dec_;
         dec->alpha_data_ += offset;
-        if (alph_dec != NULL) {
+        if (alph_dec != NULL && alph_dec->vp8l_dec_ != NULL) {
           if (alph_dec->method_ == ALPHA_LOSSLESS_COMPRESSION) {
             VP8LDecoder* const alph_vp8l_dec = alph_dec->vp8l_dec_;
-            assert(alph_vp8l_dec != NULL);
             assert(dec->alpha_data_size_ >= ALPHA_HEADER_LEN);
             VP8LBitReaderSetBuffer(&alph_vp8l_dec->br_,
                                    dec->alpha_data_ + ALPHA_HEADER_LEN,
@@ -474,6 +473,12 @@ static VP8StatusCode DecodeRemaining(WebPIDecoder* const idec) {
             MemDataSize(&idec->mem_) > MAX_MB_SIZE) {
           return IDecError(idec, VP8_STATUS_BITSTREAM_ERROR);
         }
+        // Synchronize the threads.
+        if (dec->mt_method_ > 0) {
+          if (!WebPGetWorkerInterface()->Sync(&dec->worker_)) {
+            return IDecError(idec, VP8_STATUS_BITSTREAM_ERROR);
+          }
+        }
         RestoreContext(&context, dec, token_br);
         return VP8_STATUS_SUSPENDED;
       }

src/dec/vp8_dec.c

@@ -491,7 +491,7 @@ static int GetCoeffsAlt(VP8BitReader* const br,
   return 16;
 }
 
-WEBP_TSAN_IGNORE_FUNCTION static void InitGetCoeffs(void) {
+static WEBP_TSAN_IGNORE_FUNCTION void InitGetCoeffs(void) {
   if (GetCoeffs == NULL) {
     if (VP8GetCPUInfo != NULL && VP8GetCPUInfo(kSlowSSSE3)) {
       GetCoeffs = GetCoeffsAlt;

src/dec/vp8_dec.h

@@ -182,4 +182,4 @@ WEBP_EXTERN int VP8LGetInfo(
 }    // extern "C"
 #endif
 
-#endif  /* WEBP_DEC_VP8_DEC_H_ */
+#endif  // WEBP_DEC_VP8_DEC_H_

src/dec/vp8i_dec.h

@@ -30,8 +30,8 @@ extern "C" {
 // Various defines and enums
 
 // version numbers
-#define DEC_MAJ_VERSION 0
-#define DEC_MIN_VERSION 6
+#define DEC_MAJ_VERSION 1
+#define DEC_MIN_VERSION 0
 #define DEC_REV_VERSION 1
 
 // YUV-cache parameters. Cache is 32-bytes wide (= one cacheline).
@@ -316,4 +316,4 @@ const uint8_t* VP8DecompressAlphaRows(VP8Decoder* const dec,
 }    // extern "C"
 #endif
 
-#endif  /* WEBP_DEC_VP8I_DEC_H_ */
+#endif  // WEBP_DEC_VP8I_DEC_H_

src/dec/vp8l_dec.c

@@ -253,11 +253,11 @@ static int ReadHuffmanCodeLengths(
   int symbol;
   int max_symbol;
   int prev_code_len = DEFAULT_CODE_LENGTH;
-  HuffmanTables tables;
+  HuffmanCode table[1 << LENGTHS_TABLE_BITS];
 
-  if (!VP8LHuffmanTablesAllocate(1 << LENGTHS_TABLE_BITS, &tables) ||
-      !VP8LBuildHuffmanTable(&tables, LENGTHS_TABLE_BITS,
-                             code_length_code_lengths, NUM_CODE_LENGTH_CODES)) {
+  if (!VP8LBuildHuffmanTable(table, LENGTHS_TABLE_BITS,
+                             code_length_code_lengths,
+                             NUM_CODE_LENGTH_CODES)) {
     goto End;
   }
 
@@ -277,7 +277,7 @@ static int ReadHuffmanCodeLengths(
     int code_len;
     if (max_symbol-- == 0) break;
     VP8LFillBitWindow(br);
-    p = &tables.curr_segment->start[VP8LPrefetchBits(br) & LENGTHS_TABLE_MASK];
+    p = &table[VP8LPrefetchBits(br) & LENGTHS_TABLE_MASK];
     VP8LSetBitPos(br, br->bit_pos_ + p->bits);
     code_len = p->value;
     if (code_len < kCodeLengthLiterals) {
@@ -300,7 +300,6 @@ static int ReadHuffmanCodeLengths(
   ok = 1;
 
  End:
-  VP8LHuffmanTablesDeallocate(&tables);
   if (!ok) dec->status_ = VP8_STATUS_BITSTREAM_ERROR;
   return ok;
 }
@@ -308,8 +307,7 @@ static int ReadHuffmanCodeLengths(
 // 'code_lengths' is pre-allocated temporary buffer, used for creating Huffman
 // tree.
 static int ReadHuffmanCode(int alphabet_size, VP8LDecoder* const dec,
-                           int* const code_lengths,
-                           HuffmanTables* const table) {
+                           int* const code_lengths, HuffmanCode* const table) {
   int ok = 0;
   int size = 0;
   VP8LBitReader* const br = &dec->br_;
@@ -364,7 +362,12 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize,
   VP8LMetadata* const hdr = &dec->hdr_;
   uint32_t* huffman_image = NULL;
   HTreeGroup* htree_groups = NULL;
-  HuffmanTables* huffman_tables = &hdr->huffman_tables_;
+  // When reading htrees, some might be unused, as the format allows it.
+  // We will still read them but put them in this htree_group_bogus.
+  HTreeGroup htree_group_bogus;
+  HuffmanCode* huffman_tables = NULL;
+  HuffmanCode* huffman_tables_bogus = NULL;
+  HuffmanCode* next = NULL;
   int num_htree_groups = 1;
   int num_htree_groups_max = 1;
   int max_alphabet_size = 0;
@@ -373,10 +376,6 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize,
   int* mapping = NULL;
   int ok = 0;
 
-  // Check the table has been 0 initialized (through InitMetadata).
-  assert(huffman_tables->root.start == NULL);
-  assert(huffman_tables->curr_segment == NULL);
-
   if (allow_recursion && VP8LReadBits(br, 1)) {
     // use meta Huffman codes.
     const int huffman_precision = VP8LReadBits(br, 3) + 2;
@@ -419,6 +418,12 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize,
         if (*mapped_group == -1) *mapped_group = num_htree_groups++;
         huffman_image[i] = *mapped_group;
       }
+      huffman_tables_bogus = (HuffmanCode*)WebPSafeMalloc(
+          table_size, sizeof(*huffman_tables_bogus));
+      if (huffman_tables_bogus == NULL) {
+        dec->status_ = VP8_STATUS_OUT_OF_MEMORY;
+        goto Error;
+      }
     } else {
       num_htree_groups = num_htree_groups_max;
     }
@@ -437,82 +442,74 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize,
     }
   }
 
-  htree_groups = VP8LHtreeGroupsNew(num_htree_groups);
   code_lengths = (int*)WebPSafeCalloc((uint64_t)max_alphabet_size,
                                       sizeof(*code_lengths));
+  huffman_tables = (HuffmanCode*)WebPSafeMalloc(num_htree_groups * table_size,
+                                                sizeof(*huffman_tables));
+  htree_groups = VP8LHtreeGroupsNew(num_htree_groups);
 
-  if (htree_groups == NULL || code_lengths == NULL ||
-      !VP8LHuffmanTablesAllocate(num_htree_groups * table_size,
-                                 huffman_tables)) {
+  if (htree_groups == NULL || code_lengths == NULL || huffman_tables == NULL) {
     dec->status_ = VP8_STATUS_OUT_OF_MEMORY;
     goto Error;
   }
 
+  next = huffman_tables;
   for (i = 0; i < num_htree_groups_max; ++i) {
-    // If the index "i" is unused in the Huffman image, just make sure the
-    // coefficients are valid but do not store them.
-    if (mapping != NULL && mapping[i] == -1) {
-      for (j = 0; j < HUFFMAN_CODES_PER_META_CODE; ++j) {
-        int alphabet_size = kAlphabetSize[j];
-        if (j == 0 && color_cache_bits > 0) {
-          alphabet_size += (1 << color_cache_bits);
-        }
-        // Passing in NULL so that nothing gets filled.
-        if (!ReadHuffmanCode(alphabet_size, dec, code_lengths, NULL)) {
-          goto Error;
-        }
+    // If the index "i" is unused in the Huffman image, read the coefficients
+    // but store them to a bogus htree_group.
+    const int is_bogus = (mapping != NULL && mapping[i] == -1);
+    HTreeGroup* const htree_group =
+        is_bogus ? &htree_group_bogus :
+        &htree_groups[(mapping == NULL) ? i : mapping[i]];
+    HuffmanCode** const htrees = htree_group->htrees;
+    HuffmanCode* huffman_tables_i = is_bogus ? huffman_tables_bogus : next;
+    int size;
+    int total_size = 0;
+    int is_trivial_literal = 1;
+    int max_bits = 0;
+    for (j = 0; j < HUFFMAN_CODES_PER_META_CODE; ++j) {
+      int alphabet_size = kAlphabetSize[j];
+      htrees[j] = huffman_tables_i;
+      if (j == 0 && color_cache_bits > 0) {
+        alphabet_size += 1 << color_cache_bits;
       }
-    } else {
-      HTreeGroup* const htree_group =
-          &htree_groups[(mapping == NULL) ? i : mapping[i]];
-      HuffmanCode** const htrees = htree_group->htrees;
-      int size;
-      int total_size = 0;
-      int is_trivial_literal = 1;
-      int max_bits = 0;
-      for (j = 0; j < HUFFMAN_CODES_PER_META_CODE; ++j) {
-        int alphabet_size = kAlphabetSize[j];
-        if (j == 0 && color_cache_bits > 0) {
-          alphabet_size += (1 << color_cache_bits);
-        }
-        size =
-            ReadHuffmanCode(alphabet_size, dec, code_lengths, huffman_tables);
-        htrees[j] = huffman_tables->curr_segment->curr_table;
-        if (size == 0) {
-          goto Error;
-        }
-        if (is_trivial_literal && kLiteralMap[j] == 1) {
-          is_trivial_literal = (htrees[j]->bits == 0);
-        }
-        total_size += htrees[j]->bits;
-        huffman_tables->curr_segment->curr_table += size;
-        if (j <= ALPHA) {
-          int local_max_bits = code_lengths[0];
-          int k;
-          for (k = 1; k < alphabet_size; ++k) {
-            if (code_lengths[k] > local_max_bits) {
-              local_max_bits = code_lengths[k];
-            }
+      size =
+          ReadHuffmanCode(alphabet_size, dec, code_lengths, huffman_tables_i);
+      if (size == 0) {
+        goto Error;
+      }
+      if (is_trivial_literal && kLiteralMap[j] == 1) {
+        is_trivial_literal = (huffman_tables_i->bits == 0);
+      }
+      total_size += huffman_tables_i->bits;
+      huffman_tables_i += size;
+      if (j <= ALPHA) {
+        int local_max_bits = code_lengths[0];
+        int k;
+        for (k = 1; k < alphabet_size; ++k) {
+          if (code_lengths[k] > local_max_bits) {
+            local_max_bits = code_lengths[k];
           }
-          max_bits += local_max_bits;
         }
+        max_bits += local_max_bits;
       }
-      htree_group->is_trivial_literal = is_trivial_literal;
-      htree_group->is_trivial_code = 0;
-      if (is_trivial_literal) {
-        const int red = htrees[RED][0].value;
-        const int blue = htrees[BLUE][0].value;
-        const int alpha = htrees[ALPHA][0].value;
-        htree_group->literal_arb = ((uint32_t)alpha << 24) | (red << 16) | blue;
-        if (total_size == 0 && htrees[GREEN][0].value < NUM_LITERAL_CODES) {
-          htree_group->is_trivial_code = 1;
-          htree_group->literal_arb |= htrees[GREEN][0].value << 8;
-        }
-      }
-      htree_group->use_packed_table =
-          !htree_group->is_trivial_code && (max_bits < HUFFMAN_PACKED_BITS);
-      if (htree_group->use_packed_table) BuildPackedTable(htree_group);
     }
+    if (!is_bogus) next = huffman_tables_i;
+    htree_group->is_trivial_literal = is_trivial_literal;
+    htree_group->is_trivial_code = 0;
+    if (is_trivial_literal) {
+      const int red = htrees[RED][0].value;
+      const int blue = htrees[BLUE][0].value;
+      const int alpha = htrees[ALPHA][0].value;
+      htree_group->literal_arb = ((uint32_t)alpha << 24) | (red << 16) | blue;
+      if (total_size == 0 && htrees[GREEN][0].value < NUM_LITERAL_CODES) {
+        htree_group->is_trivial_code = 1;
+        htree_group->literal_arb |= htrees[GREEN][0].value << 8;
+      }
+    }
+    htree_group->use_packed_table =
+        !htree_group->is_trivial_code && (max_bits < HUFFMAN_PACKED_BITS);
+    if (htree_group->use_packed_table) BuildPackedTable(htree_group);
   }
   ok = 1;
 
@@ -520,13 +517,15 @@ static int ReadHuffmanCodes(VP8LDecoder* const dec, int xsize, int ysize,
   hdr->huffman_image_ = huffman_image;
   hdr->num_htree_groups_ = num_htree_groups;
   hdr->htree_groups_ = htree_groups;
+  hdr->huffman_tables_ = huffman_tables;
 
  Error:
   WebPSafeFree(code_lengths);
+  WebPSafeFree(huffman_tables_bogus);
   WebPSafeFree(mapping);
   if (!ok) {
     WebPSafeFree(huffman_image);
-    VP8LHuffmanTablesDeallocate(huffman_tables);
+    WebPSafeFree(huffman_tables);
     VP8LHtreeGroupsFree(htree_groups);
   }
   return ok;
@@ -1237,20 +1236,9 @@ static int DecodeImageData(VP8LDecoder* const dec, uint32_t* const data,
   }
 
   br->eos_ = VP8LIsEndOfStream(br);
-  // In incremental decoding:
-  // br->eos_ && src < src_last: if 'br' reached the end of the buffer and
-  // 'src_last' has not been reached yet, there is not enough data. 'dec' has to
-  // be reset until there is more data.
-  // !br->eos_ && src < src_last: this cannot happen as either the buffer is
-  // fully read, either enough has been read to reach 'src_last'.
-  // src >= src_last: 'src_last' is reached, all is fine. 'src' can actually go
-  // beyond 'src_last' in case the image is cropped and an LZ77 goes further.
-  // The buffer might have been enough or there is some left. 'br->eos_' does
-  // not matter.
-  assert(!dec->incremental_ || (br->eos_ && src < src_last) || src >= src_last);
-  if (dec->incremental_ && br->eos_ && src < src_last) {
+  if (dec->incremental_ && br->eos_ && src < src_end) {
     RestoreState(dec);
-  } else if ((dec->incremental_ && src >= src_last) || !br->eos_) {
+  } else if (!br->eos_) {
     // Process the remaining rows corresponding to last row-block.
     if (process_func != NULL) {
       process_func(dec, row > last_row ? last_row : row);
@@ -1369,7 +1357,7 @@ static void ClearMetadata(VP8LMetadata* const hdr) {
   assert(hdr != NULL);
 
   WebPSafeFree(hdr->huffman_image_);
-  VP8LHuffmanTablesDeallocate(&hdr->huffman_tables_);
+  WebPSafeFree(hdr->huffman_tables_);
   VP8LHtreeGroupsFree(hdr->htree_groups_);
   VP8LColorCacheClear(&hdr->color_cache_);
   VP8LColorCacheClear(&hdr->saved_color_cache_);
@@ -1588,7 +1576,6 @@ int VP8LDecodeAlphaHeader(ALPHDecoder* const alph_dec,
   if (dec == NULL) return 0;
 
   assert(alph_dec != NULL);
-  alph_dec->vp8l_dec_ = dec;
 
   dec->width_ = alph_dec->width_;
   dec->height_ = alph_dec->height_;
@@ -1620,11 +1607,12 @@ int VP8LDecodeAlphaHeader(ALPHDecoder* const alph_dec,
 
   if (!ok) goto Err;
 
+  // Only set here, once we are sure it is valid (to avoid thread races).
+  alph_dec->vp8l_dec_ = dec;
   return 1;
 
  Err:
-  VP8LDelete(alph_dec->vp8l_dec_);
-  alph_dec->vp8l_dec_ = NULL;
+  VP8LDelete(dec);
   return 0;
 }
 
@@ -1685,7 +1673,7 @@ int VP8LDecodeImage(VP8LDecoder* const dec) {
   // Sanity checks.
   if (dec == NULL) return 0;
 
-  assert(dec->hdr_.huffman_tables_.root.start != NULL);
+  assert(dec->hdr_.huffman_tables_ != NULL);
   assert(dec->hdr_.htree_groups_ != NULL);
   assert(dec->hdr_.num_htree_groups_ > 0);
 
@@ -1708,17 +1696,17 @@ int VP8LDecodeImage(VP8LDecoder* const dec) {
 
 #if !defined(WEBP_REDUCE_SIZE)
     if (io->use_scaling && !AllocateAndInitRescaler(dec, io)) goto Err;
-
-    if (io->use_scaling || WebPIsPremultipliedMode(dec->output_->colorspace)) {
-      // need the alpha-multiply functions for premultiplied output or rescaling
-      WebPInitAlphaProcessing();
-    }
 #else
     if (io->use_scaling) {
       dec->status_ = VP8_STATUS_INVALID_PARAM;
       goto Err;
     }
 #endif
+    if (io->use_scaling || WebPIsPremultipliedMode(dec->output_->colorspace)) {
+      // need the alpha-multiply functions for premultiplied output or rescaling
+      WebPInitAlphaProcessing();
+    }
+
     if (!WebPIsRGBMode(dec->output_->colorspace)) {
       WebPInitConvertARGBToYUV();
       if (dec->output_->u.YUVA.a != NULL) WebPInitAlphaProcessing();

src/dec/vp8li_dec.h

@@ -51,7 +51,7 @@ typedef struct {
   uint32_t       *huffman_image_;
   int             num_htree_groups_;
   HTreeGroup     *htree_groups_;
-  HuffmanTables   huffman_tables_;
+  HuffmanCode    *huffman_tables_;
 } VP8LMetadata;
 
 typedef struct VP8LDecoder VP8LDecoder;
@@ -132,4 +132,4 @@ void VP8LDelete(VP8LDecoder* const dec);
 }    // extern "C"
 #endif
 
-#endif  /* WEBP_DEC_VP8LI_DEC_H_ */
+#endif  // WEBP_DEC_VP8LI_DEC_H_

src/dec/webpi_dec.h

@@ -130,4 +130,4 @@ int WebPAvoidSlowMemory(const WebPDecBuffer* const output,
 }    // extern "C"
 #endif
 
-#endif  /* WEBP_DEC_WEBPI_DEC_H_ */
+#endif  // WEBP_DEC_WEBPI_DEC_H_

src/demux/Makefile.am

@@ -5,11 +5,14 @@ libwebpdemux_la_SOURCES =
 libwebpdemux_la_SOURCES += anim_decode.c demux.c
 
 libwebpdemuxinclude_HEADERS =
+libwebpdemuxinclude_HEADERS += ../webp/decode.h
 libwebpdemuxinclude_HEADERS += ../webp/demux.h
 libwebpdemuxinclude_HEADERS += ../webp/mux_types.h
 libwebpdemuxinclude_HEADERS += ../webp/types.h
+noinst_HEADERS =
+noinst_HEADERS += ../webp/format_constants.h
 
 libwebpdemux_la_LIBADD = ../libwebp.la
-libwebpdemux_la_LDFLAGS = -no-undefined -version-info 2:3:0
+libwebpdemux_la_LDFLAGS = -no-undefined -version-info 2:5:0
 libwebpdemuxincludedir = $(includedir)/webp
 pkgconfig_DATA = libwebpdemux.pc

src/demux/demux.c

@@ -23,9 +23,9 @@
 #include "src/webp/demux.h"
 #include "src/webp/format_constants.h"
 
-#define DMUX_MAJ_VERSION 0
-#define DMUX_MIN_VERSION 3
-#define DMUX_REV_VERSION 3
+#define DMUX_MAJ_VERSION 1
+#define DMUX_MIN_VERSION 0
+#define DMUX_REV_VERSION 1
 
 typedef struct {
   size_t start_;        // start location of the data

src/demux/libwebpdemux.rc

@@ -6,8 +6,8 @@
 LANGUAGE LANG_ENGLISH, SUBLANG_ENGLISH_US
 
 VS_VERSION_INFO VERSIONINFO
- FILEVERSION 0,3,0,3
- PRODUCTVERSION 0,3,0,3
+ FILEVERSION 1,0,0,1
+ PRODUCTVERSION 1,0,0,1
  FILEFLAGSMASK 0x3fL
 #ifdef _DEBUG
  FILEFLAGS 0x1L
@@ -24,12 +24,12 @@ BEGIN
         BEGIN
             VALUE "CompanyName", "Google, Inc."
             VALUE "FileDescription", "libwebpdemux DLL"
-            VALUE "FileVersion", "0.3.3"
+            VALUE "FileVersion", "1.0.1"
             VALUE "InternalName", "libwebpdemux.dll"
-            VALUE "LegalCopyright", "Copyright (C) 2017"
+            VALUE "LegalCopyright", "Copyright (C) 2018"
             VALUE "OriginalFilename", "libwebpdemux.dll"
             VALUE "ProductName", "WebP Image Demuxer"
-            VALUE "ProductVersion", "0.3.3"
+            VALUE "ProductVersion", "1.0.1"
         END
     END
     BLOCK "VarFileInfo"

src/dsp/Makefile.am

@@ -1,7 +1,6 @@
 AM_CPPFLAGS += -I$(top_builddir) -I$(top_srcdir)
 noinst_LTLIBRARIES =
 noinst_LTLIBRARIES += libwebpdsp.la
-noinst_LTLIBRARIES += libwebpdsp_avx2.la
 noinst_LTLIBRARIES += libwebpdsp_sse2.la
 noinst_LTLIBRARIES += libwebpdspdecode_sse2.la
 noinst_LTLIBRARIES += libwebpdsp_sse41.la
@@ -10,6 +9,10 @@ noinst_LTLIBRARIES += libwebpdsp_neon.la
 noinst_LTLIBRARIES += libwebpdspdecode_neon.la
 noinst_LTLIBRARIES += libwebpdsp_msa.la
 noinst_LTLIBRARIES += libwebpdspdecode_msa.la
+noinst_LTLIBRARIES += libwebpdsp_mips32.la
+noinst_LTLIBRARIES += libwebpdspdecode_mips32.la
+noinst_LTLIBRARIES += libwebpdsp_mips_dsp_r2.la
+noinst_LTLIBRARIES += libwebpdspdecode_mips_dsp_r2.la
 
 if BUILD_LIBWEBPDECODER
   noinst_LTLIBRARIES += libwebpdspdecode.la
@@ -20,56 +23,36 @@ commondir = $(includedir)/webp
 
 COMMON_SOURCES =
 COMMON_SOURCES += alpha_processing.c
-COMMON_SOURCES += alpha_processing_mips_dsp_r2.c
-COMMON_SOURCES += common_sse2.h
 COMMON_SOURCES += cpu.c
 COMMON_SOURCES += dec.c
 COMMON_SOURCES += dec_clip_tables.c
-COMMON_SOURCES += dec_mips32.c
-COMMON_SOURCES += dec_mips_dsp_r2.c
 COMMON_SOURCES += dsp.h
 COMMON_SOURCES += filters.c
-COMMON_SOURCES += filters_mips_dsp_r2.c
 COMMON_SOURCES += lossless.c
 COMMON_SOURCES += lossless.h
 COMMON_SOURCES += lossless_common.h
-COMMON_SOURCES += lossless_mips_dsp_r2.c
-COMMON_SOURCES += mips_macro.h
 COMMON_SOURCES += rescaler.c
-COMMON_SOURCES += rescaler_mips32.c
-COMMON_SOURCES += rescaler_mips_dsp_r2.c
 COMMON_SOURCES += upsampling.c
-COMMON_SOURCES += upsampling_mips_dsp_r2.c
 COMMON_SOURCES += yuv.c
 COMMON_SOURCES += yuv.h
-COMMON_SOURCES += yuv_mips32.c
-COMMON_SOURCES += yuv_mips_dsp_r2.c
 
 ENC_SOURCES =
 ENC_SOURCES += cost.c
-ENC_SOURCES += cost_mips32.c
-ENC_SOURCES += cost_mips_dsp_r2.c
 ENC_SOURCES += enc.c
-ENC_SOURCES += enc_mips32.c
-ENC_SOURCES += enc_mips_dsp_r2.c
 ENC_SOURCES += lossless_enc.c
-ENC_SOURCES += lossless_enc_mips32.c
-ENC_SOURCES += lossless_enc_mips_dsp_r2.c
 ENC_SOURCES += ssim.c
 
-libwebpdsp_avx2_la_SOURCES =
-libwebpdsp_avx2_la_SOURCES += enc_avx2.c
-libwebpdsp_avx2_la_CPPFLAGS = $(libwebpdsp_la_CPPFLAGS)
-libwebpdsp_avx2_la_CFLAGS = $(AM_CFLAGS) $(AVX2_FLAGS)
-
 libwebpdspdecode_sse41_la_SOURCES =
 libwebpdspdecode_sse41_la_SOURCES += alpha_processing_sse41.c
 libwebpdspdecode_sse41_la_SOURCES += dec_sse41.c
+libwebpdspdecode_sse41_la_SOURCES += upsampling_sse41.c
+libwebpdspdecode_sse41_la_SOURCES += yuv_sse41.c
 libwebpdspdecode_sse41_la_CPPFLAGS = $(libwebpdsp_la_CPPFLAGS)
 libwebpdspdecode_sse41_la_CFLAGS = $(AM_CFLAGS) $(SSE41_FLAGS)
 
 libwebpdspdecode_sse2_la_SOURCES =
 libwebpdspdecode_sse2_la_SOURCES += alpha_processing_sse2.c
+libwebpdspdecode_sse2_la_SOURCES += common_sse2.h
 libwebpdspdecode_sse2_la_SOURCES += dec_sse2.c
 libwebpdspdecode_sse2_la_SOURCES += filters_sse2.c
 libwebpdspdecode_sse2_la_SOURCES += lossless_sse2.c
@@ -101,6 +84,26 @@ libwebpdspdecode_msa_la_SOURCES += upsampling_msa.c
 libwebpdspdecode_msa_la_CPPFLAGS = $(libwebpdsp_msa_la_CPPFLAGS)
 libwebpdspdecode_msa_la_CFLAGS = $(libwebpdsp_msa_la_CFLAGS)
 
+libwebpdspdecode_mips32_la_SOURCES =
+libwebpdspdecode_mips32_la_SOURCES += dec_mips32.c
+libwebpdspdecode_mips32_la_SOURCES += mips_macro.h
+libwebpdspdecode_mips32_la_SOURCES += rescaler_mips32.c
+libwebpdspdecode_mips32_la_SOURCES += yuv_mips32.c
+libwebpdspdecode_mips32_la_CPPFLAGS = $(libwebpdsp_mips32_la_CPPFLAGS)
+libwebpdspdecode_mips32_la_CFLAGS = $(libwebpdsp_mips32_la_CFLAGS)
+
+libwebpdspdecode_mips_dsp_r2_la_SOURCES =
+libwebpdspdecode_mips_dsp_r2_la_SOURCES += alpha_processing_mips_dsp_r2.c
+libwebpdspdecode_mips_dsp_r2_la_SOURCES += dec_mips_dsp_r2.c
+libwebpdspdecode_mips_dsp_r2_la_SOURCES += filters_mips_dsp_r2.c
+libwebpdspdecode_mips_dsp_r2_la_SOURCES += lossless_mips_dsp_r2.c
+libwebpdspdecode_mips_dsp_r2_la_SOURCES += mips_macro.h
+libwebpdspdecode_mips_dsp_r2_la_SOURCES += rescaler_mips_dsp_r2.c
+libwebpdspdecode_mips_dsp_r2_la_SOURCES += upsampling_mips_dsp_r2.c
+libwebpdspdecode_mips_dsp_r2_la_SOURCES += yuv_mips_dsp_r2.c
+libwebpdspdecode_mips_dsp_r2_la_CPPFLAGS = $(libwebpdsp_mips_dsp_r2_la_CPPFLAGS)
+libwebpdspdecode_mips_dsp_r2_la_CFLAGS = $(libwebpdsp_mips_dsp_r2_la_CFLAGS)
+
 libwebpdsp_sse2_la_SOURCES =
 libwebpdsp_sse2_la_SOURCES += cost_sse2.c
 libwebpdsp_sse2_la_SOURCES += enc_sse2.c
@@ -131,6 +134,22 @@ libwebpdsp_msa_la_CPPFLAGS = $(libwebpdsp_la_CPPFLAGS)
 libwebpdsp_msa_la_CFLAGS = $(AM_CFLAGS)
 libwebpdsp_msa_la_LIBADD = libwebpdspdecode_msa.la
 
+libwebpdsp_mips32_la_SOURCES =
+libwebpdsp_mips32_la_SOURCES += cost_mips32.c
+libwebpdsp_mips32_la_SOURCES += enc_mips32.c
+libwebpdsp_mips32_la_SOURCES += lossless_enc_mips32.c
+libwebpdsp_mips32_la_CPPFLAGS = $(libwebpdsp_la_CPPFLAGS)
+libwebpdsp_mips32_la_CFLAGS = $(AM_CFLAGS)
+libwebpdsp_mips32_la_LIBADD = libwebpdspdecode_mips32.la
+
+libwebpdsp_mips_dsp_r2_la_SOURCES =
+libwebpdsp_mips_dsp_r2_la_SOURCES += cost_mips_dsp_r2.c
+libwebpdsp_mips_dsp_r2_la_SOURCES += enc_mips_dsp_r2.c
+libwebpdsp_mips_dsp_r2_la_SOURCES += lossless_enc_mips_dsp_r2.c
+libwebpdsp_mips_dsp_r2_la_CPPFLAGS = $(libwebpdsp_la_CPPFLAGS)
+libwebpdsp_mips_dsp_r2_la_CFLAGS = $(AM_CFLAGS)
+libwebpdsp_mips_dsp_r2_la_LIBADD = libwebpdspdecode_mips_dsp_r2.la
+
 libwebpdsp_la_SOURCES = $(COMMON_SOURCES) $(ENC_SOURCES)
 
 noinst_HEADERS =
@@ -139,14 +158,15 @@ noinst_HEADERS += ../webp/decode.h
 
 libwebpdsp_la_CPPFLAGS =
 libwebpdsp_la_CPPFLAGS += $(AM_CPPFLAGS)
-libwebpdsp_la_CPPFLAGS += $(USE_EXPERIMENTAL_CODE) $(USE_SWAP_16BIT_CSP)
+libwebpdsp_la_CPPFLAGS += $(USE_SWAP_16BIT_CSP)
 libwebpdsp_la_LDFLAGS = -lm
 libwebpdsp_la_LIBADD =
-libwebpdsp_la_LIBADD += libwebpdsp_avx2.la
 libwebpdsp_la_LIBADD += libwebpdsp_sse2.la
 libwebpdsp_la_LIBADD += libwebpdsp_sse41.la
 libwebpdsp_la_LIBADD += libwebpdsp_neon.la
 libwebpdsp_la_LIBADD += libwebpdsp_msa.la
+libwebpdsp_la_LIBADD += libwebpdsp_mips32.la
+libwebpdsp_la_LIBADD += libwebpdsp_mips_dsp_r2.la
 
 if BUILD_LIBWEBPDECODER
   libwebpdspdecode_la_SOURCES = $(COMMON_SOURCES)
@@ -158,4 +178,6 @@ if BUILD_LIBWEBPDECODER
   libwebpdspdecode_la_LIBADD += libwebpdspdecode_sse41.la
   libwebpdspdecode_la_LIBADD += libwebpdspdecode_neon.la
   libwebpdspdecode_la_LIBADD += libwebpdspdecode_msa.la
+  libwebpdspdecode_la_LIBADD += libwebpdspdecode_mips32.la
+  libwebpdspdecode_la_LIBADD += libwebpdspdecode_mips_dsp_r2.la
 endif

src/dsp/alpha_processing.c

@@ -409,12 +409,7 @@ extern void WebPInitAlphaProcessingSSE2(void);
 extern void WebPInitAlphaProcessingSSE41(void);
 extern void WebPInitAlphaProcessingNEON(void);
 
-static volatile VP8CPUInfo alpha_processing_last_cpuinfo_used =
-    (VP8CPUInfo)&alpha_processing_last_cpuinfo_used;
-
-WEBP_TSAN_IGNORE_FUNCTION void WebPInitAlphaProcessing(void) {
-  if (alpha_processing_last_cpuinfo_used == VP8GetCPUInfo) return;
-
+WEBP_DSP_INIT_FUNC(WebPInitAlphaProcessing) {
   WebPMultARGBRow = WebPMultARGBRow_C;
   WebPMultRow = WebPMultRow_C;
   WebPApplyAlphaMultiply4444 = ApplyAlphaMultiply_16b_C;
@@ -474,6 +469,4 @@ WEBP_TSAN_IGNORE_FUNCTION void WebPInitAlphaProcessing(void) {
   assert(WebPPackRGB != NULL);
   assert(WebPHasAlpha8b != NULL);
   assert(WebPHasAlpha32b != NULL);
-
-  alpha_processing_last_cpuinfo_used = VP8GetCPUInfo;
 }

src/dsp/alpha_processing_neon.c

@@ -83,7 +83,7 @@ static void ApplyAlphaMultiply_NEON(uint8_t* rgba, int alpha_first,
 static int DispatchAlpha_NEON(const uint8_t* alpha, int alpha_stride,
                               int width, int height,
                               uint8_t* dst, int dst_stride) {
-  uint32_t alpha_mask = 0xffu;
+  uint32_t alpha_mask = 0xffffffffu;
   uint8x8_t mask8 = vdup_n_u8(0xff);
   uint32_t tmp[2];
   int i, j;
@@ -107,7 +107,6 @@ static int DispatchAlpha_NEON(const uint8_t* alpha, int alpha_stride,
     dst += dst_stride;
   }
   vst1_u8((uint8_t*)tmp, mask8);
-  alpha_mask *= 0x01010101;
   alpha_mask &= tmp[0];
   alpha_mask &= tmp[1];
   return (alpha_mask != 0xffffffffu);
@@ -135,7 +134,7 @@ static void DispatchAlphaToGreen_NEON(const uint8_t* alpha, int alpha_stride,
 static int ExtractAlpha_NEON(const uint8_t* argb, int argb_stride,
                              int width, int height,
                              uint8_t* alpha, int alpha_stride) {
-  uint32_t alpha_mask = 0xffu;
+  uint32_t alpha_mask = 0xffffffffu;
   uint8x8_t mask8 = vdup_n_u8(0xff);
   uint32_t tmp[2];
   int i, j;
@@ -157,7 +156,6 @@ static int ExtractAlpha_NEON(const uint8_t* argb, int argb_stride,
     alpha += alpha_stride;
   }
   vst1_u8((uint8_t*)tmp, mask8);
-  alpha_mask *= 0x01010101;
   alpha_mask &= tmp[0];
   alpha_mask &= tmp[1];
   return (alpha_mask == 0xffffffffu);

src/dsp/common_sse2.h

@@ -128,9 +128,9 @@ static WEBP_INLINE void VP8Transpose_2_4x4_16b(
 // Pack the planar buffers
 // rrrr... rrrr... gggg... gggg... bbbb... bbbb....
 // triplet by triplet in the output buffer rgb as rgbrgbrgbrgb ...
-static WEBP_INLINE void VP8PlanarTo24b(__m128i* const in0, __m128i* const in1,
-                                       __m128i* const in2, __m128i* const in3,
-                                       __m128i* const in4, __m128i* const in5) {
+static WEBP_INLINE void VP8PlanarTo24b_SSE2(
+    __m128i* const in0, __m128i* const in1, __m128i* const in2,
+    __m128i* const in3, __m128i* const in4, __m128i* const in5) {
   // The input is 6 registers of sixteen 8b but for the sake of explanation,
   // let's take 6 registers of four 8b values.
   // To pack, we will keep taking one every two 8b integer and move it
@@ -159,10 +159,10 @@ static WEBP_INLINE void VP8PlanarTo24b(__m128i* const in0, __m128i* const in1,
 
 // Convert four packed four-channel buffers like argbargbargbargb... into the
 // split channels aaaaa ... rrrr ... gggg .... bbbbb ......
-static WEBP_INLINE void VP8L32bToPlanar(__m128i* const in0,
-                                        __m128i* const in1,
-                                        __m128i* const in2,
-                                        __m128i* const in3) {
+static WEBP_INLINE void VP8L32bToPlanar_SSE2(__m128i* const in0,
+                                             __m128i* const in1,
+                                             __m128i* const in2,
+                                             __m128i* const in3) {
   // Column-wise transpose.
   const __m128i A0 = _mm_unpacklo_epi8(*in0, *in1);
   const __m128i A1 = _mm_unpackhi_epi8(*in0, *in1);

src/dsp/common_sse41.h

@@ -0,0 +1,132 @@
+// Copyright 2016 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// SSE4 code common to several files.
+//
+// Author: Vincent Rabaud (vrabaud@google.com)
+
+#ifndef WEBP_DSP_COMMON_SSE41_H_
+#define WEBP_DSP_COMMON_SSE41_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#if defined(WEBP_USE_SSE41)
+#include <smmintrin.h>
+
+//------------------------------------------------------------------------------
+// Channel mixing.
+// Shuffles the input buffer as A0 0 0 A1 0 0 A2 ...
+#define WEBP_SSE41_SHUFF(OUT, IN0, IN1)    \
+  OUT##0 = _mm_shuffle_epi8(*IN0, shuff0); \
+  OUT##1 = _mm_shuffle_epi8(*IN0, shuff1); \
+  OUT##2 = _mm_shuffle_epi8(*IN0, shuff2); \
+  OUT##3 = _mm_shuffle_epi8(*IN1, shuff0); \
+  OUT##4 = _mm_shuffle_epi8(*IN1, shuff1); \
+  OUT##5 = _mm_shuffle_epi8(*IN1, shuff2);
+
+// Pack the planar buffers
+// rrrr... rrrr... gggg... gggg... bbbb... bbbb....
+// triplet by triplet in the output buffer rgb as rgbrgbrgbrgb ...
+static WEBP_INLINE void VP8PlanarTo24b_SSE41(
+    __m128i* const in0, __m128i* const in1, __m128i* const in2,
+    __m128i* const in3, __m128i* const in4, __m128i* const in5) {
+  __m128i R0, R1, R2, R3, R4, R5;
+  __m128i G0, G1, G2, G3, G4, G5;
+  __m128i B0, B1, B2, B3, B4, B5;
+
+  // Process R.
+  {
+    const __m128i shuff0 = _mm_set_epi8(
+        5, -1, -1, 4, -1, -1, 3, -1, -1, 2, -1, -1, 1, -1, -1, 0);
+    const __m128i shuff1 = _mm_set_epi8(
+        -1, 10, -1, -1, 9, -1, -1, 8, -1, -1, 7, -1, -1, 6, -1, -1);
+    const __m128i shuff2 = _mm_set_epi8(
+     -1, -1, 15, -1, -1, 14, -1, -1, 13, -1, -1, 12, -1, -1, 11, -1);
+    WEBP_SSE41_SHUFF(R, in0, in1)
+  }
+
+  // Process G.
+  {
+    // Same as before, just shifted to the left by one and including the right
+    // padding.
+    const __m128i shuff0 = _mm_set_epi8(
+        -1, -1, 4, -1, -1, 3, -1, -1, 2, -1, -1, 1, -1, -1, 0, -1);
+    const __m128i shuff1 = _mm_set_epi8(
+        10, -1, -1, 9, -1, -1, 8, -1, -1, 7, -1, -1, 6, -1, -1, 5);
+    const __m128i shuff2 = _mm_set_epi8(
+     -1, 15, -1, -1, 14, -1, -1, 13, -1, -1, 12, -1, -1, 11, -1, -1);
+    WEBP_SSE41_SHUFF(G, in2, in3)
+  }
+
+  // Process B.
+  {
+    const __m128i shuff0 = _mm_set_epi8(
+        -1, 4, -1, -1, 3, -1, -1, 2, -1, -1, 1, -1, -1, 0, -1, -1);
+    const __m128i shuff1 = _mm_set_epi8(
+        -1, -1, 9, -1, -1, 8, -1, -1, 7, -1, -1, 6, -1, -1, 5, -1);
+    const __m128i shuff2 = _mm_set_epi8(
+      15, -1, -1, 14, -1, -1, 13, -1, -1, 12, -1, -1, 11, -1, -1, 10);
+    WEBP_SSE41_SHUFF(B, in4, in5)
+  }
+
+  // OR the different channels.
+  {
+    const __m128i RG0 = _mm_or_si128(R0, G0);
+    const __m128i RG1 = _mm_or_si128(R1, G1);
+    const __m128i RG2 = _mm_or_si128(R2, G2);
+    const __m128i RG3 = _mm_or_si128(R3, G3);
+    const __m128i RG4 = _mm_or_si128(R4, G4);
+    const __m128i RG5 = _mm_or_si128(R5, G5);
+    *in0 = _mm_or_si128(RG0, B0);
+    *in1 = _mm_or_si128(RG1, B1);
+    *in2 = _mm_or_si128(RG2, B2);
+    *in3 = _mm_or_si128(RG3, B3);
+    *in4 = _mm_or_si128(RG4, B4);
+    *in5 = _mm_or_si128(RG5, B5);
+  }
+}
+
+#undef WEBP_SSE41_SHUFF
+
+// Convert four packed four-channel buffers like argbargbargbargb... into the
+// split channels aaaaa ... rrrr ... gggg .... bbbbb ......
+static WEBP_INLINE void VP8L32bToPlanar_SSE41(__m128i* const in0,
+                                              __m128i* const in1,
+                                              __m128i* const in2,
+                                              __m128i* const in3) {
+  // aaaarrrrggggbbbb
+  const __m128i shuff0 =
+      _mm_set_epi8(15, 11, 7, 3, 14, 10, 6, 2, 13, 9, 5, 1, 12, 8, 4, 0);
+  const __m128i A0 = _mm_shuffle_epi8(*in0, shuff0);
+  const __m128i A1 = _mm_shuffle_epi8(*in1, shuff0);
+  const __m128i A2 = _mm_shuffle_epi8(*in2, shuff0);
+  const __m128i A3 = _mm_shuffle_epi8(*in3, shuff0);
+  // A0A1R0R1
+  // G0G1B0B1
+  // A2A3R2R3
+  // G0G1B0B1
+  const __m128i B0 = _mm_unpacklo_epi32(A0, A1);
+  const __m128i B1 = _mm_unpackhi_epi32(A0, A1);
+  const __m128i B2 = _mm_unpacklo_epi32(A2, A3);
+  const __m128i B3 = _mm_unpackhi_epi32(A2, A3);
+  *in3 = _mm_unpacklo_epi64(B0, B2);
+  *in2 = _mm_unpackhi_epi64(B0, B2);
+  *in1 = _mm_unpacklo_epi64(B1, B3);
+  *in0 = _mm_unpackhi_epi64(B1, B3);
+}
+
+#endif  // WEBP_USE_SSE41
+
+#ifdef __cplusplus
+}  // extern "C"
+#endif
+
+#endif  // WEBP_DSP_COMMON_SSE41_H_

src/dsp/cost.c

@@ -378,12 +378,7 @@ extern void VP8EncDspCostInitMIPS32(void);
 extern void VP8EncDspCostInitMIPSdspR2(void);
 extern void VP8EncDspCostInitSSE2(void);
 
-static volatile VP8CPUInfo cost_last_cpuinfo_used =
-    (VP8CPUInfo)&cost_last_cpuinfo_used;
-
-WEBP_TSAN_IGNORE_FUNCTION void VP8EncDspCostInit(void) {
-  if (cost_last_cpuinfo_used == VP8GetCPUInfo) return;
-
+WEBP_DSP_INIT_FUNC(VP8EncDspCostInit) {
   VP8GetResidualCost = GetResidualCost_C;
   VP8SetResidualCoeffs = SetResidualCoeffs_C;
 
@@ -405,8 +400,6 @@ WEBP_TSAN_IGNORE_FUNCTION void VP8EncDspCostInit(void) {
     }
 #endif
   }
-
-  cost_last_cpuinfo_used = VP8GetCPUInfo;
 }
 
 //------------------------------------------------------------------------------

src/dsp/dec.c

@@ -741,12 +741,7 @@ extern void VP8DspInitMIPS32(void);
 extern void VP8DspInitMIPSdspR2(void);
 extern void VP8DspInitMSA(void);
 
-static volatile VP8CPUInfo dec_last_cpuinfo_used =
-    (VP8CPUInfo)&dec_last_cpuinfo_used;
-
-WEBP_TSAN_IGNORE_FUNCTION void VP8DspInit(void) {
-  if (dec_last_cpuinfo_used == VP8GetCPUInfo) return;
-
+WEBP_DSP_INIT_FUNC(VP8DspInit) {
   VP8InitClipTables();
 
 #if !WEBP_NEON_OMIT_C_CODE
@@ -889,6 +884,4 @@ WEBP_TSAN_IGNORE_FUNCTION void VP8DspInit(void) {
   assert(VP8PredChroma8[5] != NULL);
   assert(VP8PredChroma8[6] != NULL);
   assert(VP8DitherCombine8x8 != NULL);
-
-  dec_last_cpuinfo_used = VP8GetCPUInfo;
 }

src/dsp/dsp.h

@@ -76,10 +76,6 @@ extern "C" {
 #define WEBP_USE_SSE41
 #endif
 
-#if defined(__AVX2__) || defined(WEBP_HAVE_AVX2)
-#define WEBP_USE_AVX2
-#endif
-
 // The intrinsics currently cause compiler errors with arm-nacl-gcc and the
 // inline assembly would need to be modified for use with Native Client.
 #if (defined(__ARM_NEON__) || \
@@ -141,6 +137,42 @@ extern "C" {
 #endif
 #endif
 
+#if defined(WEBP_USE_THREAD) && !defined(_WIN32)
+#include <pthread.h>  // NOLINT
+
+#define WEBP_DSP_INIT(func) do {                                    \
+  static volatile VP8CPUInfo func ## _last_cpuinfo_used =           \
+      (VP8CPUInfo)&func ## _last_cpuinfo_used;                      \
+  static pthread_mutex_t func ## _lock = PTHREAD_MUTEX_INITIALIZER; \
+  if (pthread_mutex_lock(&func ## _lock)) break;                    \
+  if (func ## _last_cpuinfo_used != VP8GetCPUInfo) func();          \
+  func ## _last_cpuinfo_used = VP8GetCPUInfo;                       \
+  (void)pthread_mutex_unlock(&func ## _lock);                       \
+} while (0)
+#else  // !(defined(WEBP_USE_THREAD) && !defined(_WIN32))
+#define WEBP_DSP_INIT(func) do {                                    \
+  static volatile VP8CPUInfo func ## _last_cpuinfo_used =           \
+      (VP8CPUInfo)&func ## _last_cpuinfo_used;                      \
+  if (func ## _last_cpuinfo_used == VP8GetCPUInfo) break;           \
+  func();                                                           \
+  func ## _last_cpuinfo_used = VP8GetCPUInfo;                       \
+} while (0)
+#endif  // defined(WEBP_USE_THREAD) && !defined(_WIN32)
+
+// Defines an Init + helper function that control multiple initialization of
+// function pointers / tables.
+/* Usage:
+   WEBP_DSP_INIT_FUNC(InitFunc) {
+     ...function body
+   }
+*/
+#define WEBP_DSP_INIT_FUNC(name)                             \
+  static WEBP_TSAN_IGNORE_FUNCTION void name ## _body(void); \
+  WEBP_TSAN_IGNORE_FUNCTION void name(void) {                \
+    WEBP_DSP_INIT(name ## _body);                            \
+  }                                                          \
+  static WEBP_TSAN_IGNORE_FUNCTION void name ## _body(void)
+
 #define WEBP_UBSAN_IGNORE_UNDEF
 #define WEBP_UBSAN_IGNORE_UNSIGNED_OVERFLOW
 #if defined(__clang__) && defined(__has_attribute)
@@ -196,7 +228,7 @@ WEBP_EXTERN VP8CPUInfo VP8GetCPUInfo;
 // avoiding a compiler warning.
 #define WEBP_DSP_INIT_STUB(func) \
   extern void func(void); \
-  WEBP_TSAN_IGNORE_FUNCTION void func(void) {}
+  void func(void) {}
 
 //------------------------------------------------------------------------------
 // Encoding
@@ -643,4 +675,4 @@ void VP8FiltersInit(void);
 }    // extern "C"
 #endif
 
-#endif  /* WEBP_DSP_DSP_H_ */
+#endif  // WEBP_DSP_DSP_H_

src/dsp/enc.c

@@ -734,18 +734,12 @@ VP8BlockCopy VP8Copy16x8;
 
 extern void VP8EncDspInitSSE2(void);
 extern void VP8EncDspInitSSE41(void);
-extern void VP8EncDspInitAVX2(void);
 extern void VP8EncDspInitNEON(void);
 extern void VP8EncDspInitMIPS32(void);
 extern void VP8EncDspInitMIPSdspR2(void);
 extern void VP8EncDspInitMSA(void);
 
-static volatile VP8CPUInfo enc_last_cpuinfo_used =
-    (VP8CPUInfo)&enc_last_cpuinfo_used;
-
-WEBP_TSAN_IGNORE_FUNCTION void VP8EncDspInit(void) {
-  if (enc_last_cpuinfo_used == VP8GetCPUInfo) return;
-
+WEBP_DSP_INIT_FUNC(VP8EncDspInit) {
   VP8DspInit();  // common inverse transforms
   InitTables();
 
@@ -789,11 +783,6 @@ WEBP_TSAN_IGNORE_FUNCTION void VP8EncDspInit(void) {
 #endif
     }
 #endif
-#if defined(WEBP_USE_AVX2)
-    if (VP8GetCPUInfo(kAVX2)) {
-      VP8EncDspInitAVX2();
-    }
-#endif
 #if defined(WEBP_USE_MIPS32)
     if (VP8GetCPUInfo(kMIPS32)) {
       VP8EncDspInitMIPS32();
@@ -838,6 +827,4 @@ WEBP_TSAN_IGNORE_FUNCTION void VP8EncDspInit(void) {
   assert(VP8EncQuantizeBlockWHT != NULL);
   assert(VP8Copy4x4 != NULL);
   assert(VP8Copy16x8 != NULL);
-
-  enc_last_cpuinfo_used = VP8GetCPUInfo;
 }

src/dsp/enc_avx2.c

@@ -1,21 +0,0 @@
-// Copyright 2014 Google Inc. All Rights Reserved.
-//
-// Use of this source code is governed by a BSD-style license
-// that can be found in the COPYING file in the root of the source
-// tree. An additional intellectual property rights grant can be found
-// in the file PATENTS. All contributing project authors may
-// be found in the AUTHORS file in the root of the source tree.
-// -----------------------------------------------------------------------------
-//
-// AVX2 version of speed-critical encoding functions.
-
-#include "src/dsp/dsp.h"
-
-#if defined(WEBP_USE_AVX2)
-
-#endif  // WEBP_USE_AVX2
-
-//------------------------------------------------------------------------------
-// Entry point
-
-WEBP_DSP_INIT_STUB(VP8EncDspInitAVX2)

src/dsp/filters.c

@@ -238,12 +238,7 @@ extern void VP8FiltersInitMSA(void);
 extern void VP8FiltersInitNEON(void);
 extern void VP8FiltersInitSSE2(void);
 
-static volatile VP8CPUInfo filters_last_cpuinfo_used =
-    (VP8CPUInfo)&filters_last_cpuinfo_used;
-
-WEBP_TSAN_IGNORE_FUNCTION void VP8FiltersInit(void) {
-  if (filters_last_cpuinfo_used == VP8GetCPUInfo) return;
-
+WEBP_DSP_INIT_FUNC(VP8FiltersInit) {
   WebPUnfilters[WEBP_FILTER_NONE] = NULL;
 #if !WEBP_NEON_OMIT_C_CODE
   WebPUnfilters[WEBP_FILTER_HORIZONTAL] = HorizontalUnfilter_C;
@@ -289,6 +284,4 @@ WEBP_TSAN_IGNORE_FUNCTION void VP8FiltersInit(void) {
   assert(WebPFilters[WEBP_FILTER_HORIZONTAL] != NULL);
   assert(WebPFilters[WEBP_FILTER_VERTICAL] != NULL);
   assert(WebPFilters[WEBP_FILTER_GRADIENT] != NULL);
-
-  filters_last_cpuinfo_used = VP8GetCPUInfo;
 }

src/dsp/lossless.c

@@ -23,8 +23,6 @@
 #include "src/dsp/lossless.h"
 #include "src/dsp/lossless_common.h"
 
-#define MAX_DIFF_COST (1e30f)
-
 //------------------------------------------------------------------------------
 // Image transforms.
 
@@ -577,9 +575,6 @@ extern void VP8LDspInitNEON(void);
 extern void VP8LDspInitMIPSdspR2(void);
 extern void VP8LDspInitMSA(void);
 
-static volatile VP8CPUInfo lossless_last_cpuinfo_used =
-    (VP8CPUInfo)&lossless_last_cpuinfo_used;
-
 #define COPY_PREDICTOR_ARRAY(IN, OUT) do {                \
   (OUT)[0] = IN##0_C;                                     \
   (OUT)[1] = IN##1_C;                                     \
@@ -599,9 +594,7 @@ static volatile VP8CPUInfo lossless_last_cpuinfo_used =
   (OUT)[15] = IN##0_C;                                    \
 } while (0);
 
-WEBP_TSAN_IGNORE_FUNCTION void VP8LDspInit(void) {
-  if (lossless_last_cpuinfo_used == VP8GetCPUInfo) return;
-
+WEBP_DSP_INIT_FUNC(VP8LDspInit) {
   COPY_PREDICTOR_ARRAY(Predictor, VP8LPredictors)
   COPY_PREDICTOR_ARRAY(Predictor, VP8LPredictors_C)
   COPY_PREDICTOR_ARRAY(PredictorAdd, VP8LPredictorsAdd)
@@ -658,8 +651,6 @@ WEBP_TSAN_IGNORE_FUNCTION void VP8LDspInit(void) {
   assert(VP8LConvertBGRAToRGB565 != NULL);
   assert(VP8LMapColor32b != NULL);
   assert(VP8LMapColor8b != NULL);
-
-  lossless_last_cpuinfo_used = VP8GetCPUInfo;
 }
 #undef COPY_PREDICTOR_ARRAY
 

src/dsp/lossless.h

@@ -25,10 +25,6 @@
 extern "C" {
 #endif
 
-#ifdef WEBP_EXPERIMENTAL_FEATURES
-#include "src/enc/delta_palettization_enc.h"
-#endif  // WEBP_EXPERIMENTAL_FEATURES
-
 //------------------------------------------------------------------------------
 // Decoding
 
@@ -167,7 +163,7 @@ extern VP8LCostCombinedFunc VP8LExtraCostCombined;
 extern VP8LCombinedShannonEntropyFunc VP8LCombinedShannonEntropy;
 
 typedef struct {        // small struct to hold counters
-  int counts[2];        // index: 0=zero steak, 1=non-zero streak
+  int counts[2];        // index: 0=zero streak, 1=non-zero streak
   int streaks[2][2];    // [zero/non-zero][streak<3 / streak>=3]
 } VP8LStreaks;
 
@@ -198,10 +194,14 @@ extern VP8LGetEntropyUnrefinedFunc VP8LGetEntropyUnrefined;
 void VP8LBitsEntropyUnrefined(const uint32_t* const array, int n,
                               VP8LBitEntropy* const entropy);
 
-typedef void (*VP8LHistogramAddFunc)(const VP8LHistogram* const a,
-                                     const VP8LHistogram* const b,
-                                     VP8LHistogram* const out);
-extern VP8LHistogramAddFunc VP8LHistogramAdd;
+typedef void (*VP8LAddVectorFunc)(const uint32_t* a, const uint32_t* b,
+                                  uint32_t* out, int size);
+extern VP8LAddVectorFunc VP8LAddVector;
+typedef void (*VP8LAddVectorEqFunc)(const uint32_t* a, uint32_t* out, int size);
+extern VP8LAddVectorEqFunc VP8LAddVectorEq;
+void VP8LHistogramAdd(const VP8LHistogram* const a,
+                      const VP8LHistogram* const b,
+                      VP8LHistogram* const out);
 
 // -----------------------------------------------------------------------------
 // PrefixEncode()

src/dsp/lossless_enc.c

@@ -632,38 +632,67 @@ static double ExtraCostCombined_C(const uint32_t* X, const uint32_t* Y,
 
 //------------------------------------------------------------------------------
 
-static void HistogramAdd_C(const VP8LHistogram* const a,
-                           const VP8LHistogram* const b,
-                           VP8LHistogram* const out) {
+static void AddVector_C(const uint32_t* a, const uint32_t* b, uint32_t* out,
+                        int size) {
+  int i;
+  for (i = 0; i < size; ++i) out[i] = a[i] + b[i];
+}
+
+static void AddVectorEq_C(const uint32_t* a, uint32_t* out, int size) {
+  int i;
+  for (i = 0; i < size; ++i) out[i] += a[i];
+}
+
+#define ADD(X, ARG, LEN) do {                                                  \
+  if (a->is_used_[X]) {                                                        \
+    if (b->is_used_[X]) {                                                      \
+      VP8LAddVector(a->ARG, b->ARG, out->ARG, (LEN));                          \
+    } else {                                                                   \
+      memcpy(&out->ARG[0], &a->ARG[0], (LEN) * sizeof(out->ARG[0]));           \
+    }                                                                          \
+  } else if (b->is_used_[X]) {                                                 \
+    memcpy(&out->ARG[0], &b->ARG[0], (LEN) * sizeof(out->ARG[0]));             \
+  } else {                                                                     \
+    memset(&out->ARG[0], 0, (LEN) * sizeof(out->ARG[0]));                      \
+  }                                                                            \
+} while (0)
+
+#define ADD_EQ(X, ARG, LEN) do {                                               \
+  if (a->is_used_[X]) {                                                        \
+    if (out->is_used_[X]) {                                                    \
+      VP8LAddVectorEq(a->ARG, out->ARG, (LEN));                                \
+    } else {                                                                   \
+      memcpy(&out->ARG[0], &a->ARG[0], (LEN) * sizeof(out->ARG[0]));           \
+    }                                                                          \
+  }                                                                            \
+} while (0)
+
+void VP8LHistogramAdd(const VP8LHistogram* const a,
+                      const VP8LHistogram* const b, VP8LHistogram* const out) {
   int i;
   const int literal_size = VP8LHistogramNumCodes(a->palette_code_bits_);
   assert(a->palette_code_bits_ == b->palette_code_bits_);
+
   if (b != out) {
-    for (i = 0; i < literal_size; ++i) {
-      out->literal_[i] = a->literal_[i] + b->literal_[i];
-    }
-    for (i = 0; i < NUM_DISTANCE_CODES; ++i) {
-      out->distance_[i] = a->distance_[i] + b->distance_[i];
-    }
-    for (i = 0; i < NUM_LITERAL_CODES; ++i) {
-      out->red_[i] = a->red_[i] + b->red_[i];
-      out->blue_[i] = a->blue_[i] + b->blue_[i];
-      out->alpha_[i] = a->alpha_[i] + b->alpha_[i];
+    ADD(0, literal_, literal_size);
+    ADD(1, red_, NUM_LITERAL_CODES);
+    ADD(2, blue_, NUM_LITERAL_CODES);
+    ADD(3, alpha_, NUM_LITERAL_CODES);
+    ADD(4, distance_, NUM_DISTANCE_CODES);
+    for (i = 0; i < 5; ++i) {
+      out->is_used_[i] = (a->is_used_[i] | b->is_used_[i]);
     }
   } else {
-    for (i = 0; i < literal_size; ++i) {
-      out->literal_[i] += a->literal_[i];
-    }
-    for (i = 0; i < NUM_DISTANCE_CODES; ++i) {
-      out->distance_[i] += a->distance_[i];
-    }
-    for (i = 0; i < NUM_LITERAL_CODES; ++i) {
-      out->red_[i] += a->red_[i];
-      out->blue_[i] += a->blue_[i];
-      out->alpha_[i] += a->alpha_[i];
-    }
+    ADD_EQ(0, literal_, literal_size);
+    ADD_EQ(1, red_, NUM_LITERAL_CODES);
+    ADD_EQ(2, blue_, NUM_LITERAL_CODES);
+    ADD_EQ(3, alpha_, NUM_LITERAL_CODES);
+    ADD_EQ(4, distance_, NUM_DISTANCE_CODES);
+    for (i = 0; i < 5; ++i) out->is_used_[i] |= a->is_used_[i];
   }
 }
+#undef ADD
+#undef ADD_EQ
 
 //------------------------------------------------------------------------------
 // Image transforms.
@@ -848,7 +877,8 @@ VP8LCombinedShannonEntropyFunc VP8LCombinedShannonEntropy;
 VP8LGetEntropyUnrefinedFunc VP8LGetEntropyUnrefined;
 VP8LGetCombinedEntropyUnrefinedFunc VP8LGetCombinedEntropyUnrefined;
 
-VP8LHistogramAddFunc VP8LHistogramAdd;
+VP8LAddVectorFunc VP8LAddVector;
+VP8LAddVectorEqFunc VP8LAddVectorEq;
 
 VP8LVectorMismatchFunc VP8LVectorMismatch;
 VP8LBundleColorMapFunc VP8LBundleColorMap;
@@ -863,12 +893,7 @@ extern void VP8LEncDspInitMIPS32(void);
 extern void VP8LEncDspInitMIPSdspR2(void);
 extern void VP8LEncDspInitMSA(void);
 
-static volatile VP8CPUInfo lossless_enc_last_cpuinfo_used =
-    (VP8CPUInfo)&lossless_enc_last_cpuinfo_used;
-
-WEBP_TSAN_IGNORE_FUNCTION void VP8LEncDspInit(void) {
-  if (lossless_enc_last_cpuinfo_used == VP8GetCPUInfo) return;
-
+WEBP_DSP_INIT_FUNC(VP8LEncDspInit) {
   VP8LDspInit();
 
 #if !WEBP_NEON_OMIT_C_CODE
@@ -890,7 +915,8 @@ WEBP_TSAN_IGNORE_FUNCTION void VP8LEncDspInit(void) {
   VP8LGetEntropyUnrefined = GetEntropyUnrefined_C;
   VP8LGetCombinedEntropyUnrefined = GetCombinedEntropyUnrefined_C;
 
-  VP8LHistogramAdd = HistogramAdd_C;
+  VP8LAddVector = AddVector_C;
+  VP8LAddVectorEq = AddVectorEq_C;
 
   VP8LVectorMismatch = VectorMismatch_C;
   VP8LBundleColorMap = VP8LBundleColorMap_C;
@@ -976,7 +1002,8 @@ WEBP_TSAN_IGNORE_FUNCTION void VP8LEncDspInit(void) {
   assert(VP8LCombinedShannonEntropy != NULL);
   assert(VP8LGetEntropyUnrefined != NULL);
   assert(VP8LGetCombinedEntropyUnrefined != NULL);
-  assert(VP8LHistogramAdd != NULL);
+  assert(VP8LAddVector != NULL);
+  assert(VP8LAddVectorEq != NULL);
   assert(VP8LVectorMismatch != NULL);
   assert(VP8LBundleColorMap != NULL);
   assert(VP8LPredictorsSub[0] != NULL);
@@ -1011,8 +1038,6 @@ WEBP_TSAN_IGNORE_FUNCTION void VP8LEncDspInit(void) {
   assert(VP8LPredictorsSub_C[13] != NULL);
   assert(VP8LPredictorsSub_C[14] != NULL);
   assert(VP8LPredictorsSub_C[15] != NULL);
-
-  lossless_enc_last_cpuinfo_used = VP8GetCPUInfo;
 }
 
 //------------------------------------------------------------------------------

src/dsp/lossless_enc_mips32.c

@@ -344,65 +344,29 @@ static void GetCombinedEntropyUnrefined_MIPS32(const uint32_t X[],
     ASM_END_COMMON_0                                    \
     ASM_END_COMMON_1
 
-#define ADD_VECTOR(A, B, OUT, SIZE, EXTRA_SIZE)  do {   \
-  const uint32_t* pa = (const uint32_t*)(A);            \
-  const uint32_t* pb = (const uint32_t*)(B);            \
-  uint32_t* pout = (uint32_t*)(OUT);                    \
-  const uint32_t* const LoopEnd = pa + (SIZE);          \
-  assert((SIZE) % 4 == 0);                              \
-  ASM_START                                             \
-  ADD_TO_OUT(0, 4, 8, 12, 1, pa, pb, pout)              \
-  ASM_END_0                                             \
-  if ((EXTRA_SIZE) > 0) {                               \
-    const int last = (EXTRA_SIZE);                      \
-    int i;                                              \
-    for (i = 0; i < last; ++i) pout[i] = pa[i] + pb[i]; \
-  }                                                     \
-} while (0)
-
-#define ADD_VECTOR_EQ(A, OUT, SIZE, EXTRA_SIZE)  do {   \
-  const uint32_t* pa = (const uint32_t*)(A);            \
-  uint32_t* pout = (uint32_t*)(OUT);                    \
-  const uint32_t* const LoopEnd = pa + (SIZE);          \
-  assert((SIZE) % 4 == 0);                              \
-  ASM_START                                             \
-  ADD_TO_OUT(0, 4, 8, 12, 0, pa, pout, pout)            \
-  ASM_END_1                                             \
-  if ((EXTRA_SIZE) > 0) {                               \
-    const int last = (EXTRA_SIZE);                      \
-    int i;                                              \
-    for (i = 0; i < last; ++i) pout[i] += pa[i];        \
-  }                                                     \
-} while (0)
-
-static void HistogramAdd_MIPS32(const VP8LHistogram* const a,
-                                const VP8LHistogram* const b,
-                                VP8LHistogram* const out) {
+static void AddVector_MIPS32(const uint32_t* pa, const uint32_t* pb,
+                             uint32_t* pout, int size) {
   uint32_t temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7;
-  const int extra_cache_size = VP8LHistogramNumCodes(a->palette_code_bits_)
-                             - (NUM_LITERAL_CODES + NUM_LENGTH_CODES);
-  assert(a->palette_code_bits_ == b->palette_code_bits_);
-
-  if (b != out) {
-    ADD_VECTOR(a->literal_, b->literal_, out->literal_,
-               NUM_LITERAL_CODES + NUM_LENGTH_CODES, extra_cache_size);
-    ADD_VECTOR(a->distance_, b->distance_, out->distance_,
-               NUM_DISTANCE_CODES, 0);
-    ADD_VECTOR(a->red_, b->red_, out->red_, NUM_LITERAL_CODES, 0);
-    ADD_VECTOR(a->blue_, b->blue_, out->blue_, NUM_LITERAL_CODES, 0);
-    ADD_VECTOR(a->alpha_, b->alpha_, out->alpha_, NUM_LITERAL_CODES, 0);
-  } else {
-    ADD_VECTOR_EQ(a->literal_, out->literal_,
-                  NUM_LITERAL_CODES + NUM_LENGTH_CODES, extra_cache_size);
-    ADD_VECTOR_EQ(a->distance_, out->distance_, NUM_DISTANCE_CODES, 0);
-    ADD_VECTOR_EQ(a->red_, out->red_, NUM_LITERAL_CODES, 0);
-    ADD_VECTOR_EQ(a->blue_, out->blue_, NUM_LITERAL_CODES, 0);
-    ADD_VECTOR_EQ(a->alpha_, out->alpha_, NUM_LITERAL_CODES, 0);
-  }
+  const uint32_t end = ((size) / 4) * 4;
+  const uint32_t* const LoopEnd = pa + end;
+  int i;
+  ASM_START
+  ADD_TO_OUT(0, 4, 8, 12, 1, pa, pb, pout)
+  ASM_END_0
+  for (i = end; i < size; ++i) pout[i] = pa[i] + pb[i];
+}
+
+static void AddVectorEq_MIPS32(const uint32_t* pa, uint32_t* pout, int size) {
+  uint32_t temp0, temp1, temp2, temp3, temp4, temp5, temp6, temp7;
+  const uint32_t end = ((size) / 4) * 4;
+  const uint32_t* const LoopEnd = pa + end;
+  int i;
+  ASM_START
+  ADD_TO_OUT(0, 4, 8, 12, 0, pa, pout, pout)
+  ASM_END_1
+  for (i = end; i < size; ++i) pout[i] += pa[i];
 }
 
-#undef ADD_VECTOR_EQ
-#undef ADD_VECTOR
 #undef ASM_END_1
 #undef ASM_END_0
 #undef ASM_END_COMMON_1
@@ -422,7 +386,8 @@ WEBP_TSAN_IGNORE_FUNCTION void VP8LEncDspInitMIPS32(void) {
   VP8LExtraCostCombined = ExtraCostCombined_MIPS32;
   VP8LGetEntropyUnrefined = GetEntropyUnrefined_MIPS32;
   VP8LGetCombinedEntropyUnrefined = GetCombinedEntropyUnrefined_MIPS32;
-  VP8LHistogramAdd = HistogramAdd_MIPS32;
+  VP8LAddVector = AddVector_MIPS32;
+  VP8LAddVectorEq = AddVectorEq_MIPS32;
 }
 
 #else  // !WEBP_USE_MIPS32

src/dsp/lossless_enc_sse2.c

@@ -46,16 +46,14 @@ static void SubtractGreenFromBlueAndRed_SSE2(uint32_t* argb_data,
 //------------------------------------------------------------------------------
 // Color Transform
 
+#define MK_CST_16(HI, LO) \
+  _mm_set1_epi32((int)(((uint32_t)(HI) << 16) | ((LO) & 0xffff)))
+
 static void TransformColor_SSE2(const VP8LMultipliers* const m,
                                 uint32_t* argb_data, int num_pixels) {
-  const __m128i mults_rb = _mm_set_epi16(
-      CST_5b(m->green_to_red_), CST_5b(m->green_to_blue_),
-      CST_5b(m->green_to_red_), CST_5b(m->green_to_blue_),
-      CST_5b(m->green_to_red_), CST_5b(m->green_to_blue_),
-      CST_5b(m->green_to_red_), CST_5b(m->green_to_blue_));
-  const __m128i mults_b2 = _mm_set_epi16(
-      CST_5b(m->red_to_blue_), 0, CST_5b(m->red_to_blue_), 0,
-      CST_5b(m->red_to_blue_), 0, CST_5b(m->red_to_blue_), 0);
+  const __m128i mults_rb = MK_CST_16(CST_5b(m->green_to_red_),
+                                     CST_5b(m->green_to_blue_));
+  const __m128i mults_b2 = MK_CST_16(CST_5b(m->red_to_blue_), 0);
   const __m128i mask_ag = _mm_set1_epi32(0xff00ff00);  // alpha-green masks
   const __m128i mask_rb = _mm_set1_epi32(0x00ff00ff);  // red-blue masks
   int i;
@@ -85,12 +83,8 @@ static void CollectColorBlueTransforms_SSE2(const uint32_t* argb, int stride,
                                             int tile_width, int tile_height,
                                             int green_to_blue, int red_to_blue,
                                             int histo[]) {
-  const __m128i mults_r = _mm_set_epi16(
-      CST_5b(red_to_blue), 0, CST_5b(red_to_blue), 0,
-      CST_5b(red_to_blue), 0, CST_5b(red_to_blue), 0);
-  const __m128i mults_g = _mm_set_epi16(
-      0, CST_5b(green_to_blue), 0, CST_5b(green_to_blue),
-      0, CST_5b(green_to_blue), 0, CST_5b(green_to_blue));
+  const __m128i mults_r = MK_CST_16(CST_5b(red_to_blue), 0);
+  const __m128i mults_g = MK_CST_16(0, CST_5b(green_to_blue));
   const __m128i mask_g = _mm_set1_epi32(0x00ff00);  // green mask
   const __m128i mask_b = _mm_set1_epi32(0x0000ff);  // blue mask
   int y;
@@ -135,9 +129,7 @@ static void CollectColorBlueTransforms_SSE2(const uint32_t* argb, int stride,
 static void CollectColorRedTransforms_SSE2(const uint32_t* argb, int stride,
                                            int tile_width, int tile_height,
                                            int green_to_red, int histo[]) {
-  const __m128i mults_g = _mm_set_epi16(
-      0, CST_5b(green_to_red), 0, CST_5b(green_to_red),
-      0, CST_5b(green_to_red), 0, CST_5b(green_to_red));
+  const __m128i mults_g = MK_CST_16(0, CST_5b(green_to_red));
   const __m128i mask_g = _mm_set1_epi32(0x00ff00);  // green mask
   const __m128i mask = _mm_set1_epi32(0xff);
 
@@ -174,15 +166,17 @@ static void CollectColorRedTransforms_SSE2(const uint32_t* argb, int stride,
   }
 }
 #undef SPAN
+#undef MK_CST_16
 
 //------------------------------------------------------------------------------
 
+// Note we are adding uint32_t's as *signed* int32's (using _mm_add_epi32). But
+// that's ok since the histogram values are less than 1<<28 (max picture size).
 #define LINE_SIZE 16    // 8 or 16
 static void AddVector_SSE2(const uint32_t* a, const uint32_t* b, uint32_t* out,
                            int size) {
   int i;
-  assert(size % LINE_SIZE == 0);
-  for (i = 0; i < size; i += LINE_SIZE) {
+  for (i = 0; i + LINE_SIZE <= size; i += LINE_SIZE) {
     const __m128i a0 = _mm_loadu_si128((const __m128i*)&a[i +  0]);
     const __m128i a1 = _mm_loadu_si128((const __m128i*)&a[i +  4]);
 #if (LINE_SIZE == 16)
@@ -202,12 +196,14 @@ static void AddVector_SSE2(const uint32_t* a, const uint32_t* b, uint32_t* out,
     _mm_storeu_si128((__m128i*)&out[i + 12], _mm_add_epi32(a3, b3));
 #endif
   }
+  for (; i < size; ++i) {
+    out[i] = a[i] + b[i];
+  }
 }
 
 static void AddVectorEq_SSE2(const uint32_t* a, uint32_t* out, int size) {
   int i;
-  assert(size % LINE_SIZE == 0);
-  for (i = 0; i < size; i += LINE_SIZE) {
+  for (i = 0; i + LINE_SIZE <= size; i += LINE_SIZE) {
     const __m128i a0 = _mm_loadu_si128((const __m128i*)&a[i +  0]);
     const __m128i a1 = _mm_loadu_si128((const __m128i*)&a[i +  4]);
 #if (LINE_SIZE == 16)
@@ -227,36 +223,12 @@ static void AddVectorEq_SSE2(const uint32_t* a, uint32_t* out, int size) {
     _mm_storeu_si128((__m128i*)&out[i + 12], _mm_add_epi32(a3, b3));
 #endif
   }
+  for (; i < size; ++i) {
+    out[i] += a[i];
+  }
 }
 #undef LINE_SIZE
 
-// Note we are adding uint32_t's as *signed* int32's (using _mm_add_epi32). But
-// that's ok since the histogram values are less than 1<<28 (max picture size).
-static void HistogramAdd_SSE2(const VP8LHistogram* const a,
-                              const VP8LHistogram* const b,
-                              VP8LHistogram* const out) {
-  int i;
-  const int literal_size = VP8LHistogramNumCodes(a->palette_code_bits_);
-  assert(a->palette_code_bits_ == b->palette_code_bits_);
-  if (b != out) {
-    AddVector_SSE2(a->literal_, b->literal_, out->literal_, NUM_LITERAL_CODES);
-    AddVector_SSE2(a->red_, b->red_, out->red_, NUM_LITERAL_CODES);
-    AddVector_SSE2(a->blue_, b->blue_, out->blue_, NUM_LITERAL_CODES);
-    AddVector_SSE2(a->alpha_, b->alpha_, out->alpha_, NUM_LITERAL_CODES);
-  } else {
-    AddVectorEq_SSE2(a->literal_, out->literal_, NUM_LITERAL_CODES);
-    AddVectorEq_SSE2(a->red_, out->red_, NUM_LITERAL_CODES);
-    AddVectorEq_SSE2(a->blue_, out->blue_, NUM_LITERAL_CODES);
-    AddVectorEq_SSE2(a->alpha_, out->alpha_, NUM_LITERAL_CODES);
-  }
-  for (i = NUM_LITERAL_CODES; i < literal_size; ++i) {
-    out->literal_[i] = a->literal_[i] + b->literal_[i];
-  }
-  for (i = 0; i < NUM_DISTANCE_CODES; ++i) {
-    out->distance_[i] = a->distance_[i] + b->distance_[i];
-  }
-}
-
 //------------------------------------------------------------------------------
 // Entropy
 
@@ -682,7 +654,8 @@ WEBP_TSAN_IGNORE_FUNCTION void VP8LEncDspInitSSE2(void) {
   VP8LTransformColor = TransformColor_SSE2;
   VP8LCollectColorBlueTransforms = CollectColorBlueTransforms_SSE2;
   VP8LCollectColorRedTransforms = CollectColorRedTransforms_SSE2;
-  VP8LHistogramAdd = HistogramAdd_SSE2;
+  VP8LAddVector = AddVector_SSE2;
+  VP8LAddVectorEq = AddVectorEq_SSE2;
   VP8LCombinedShannonEntropy = CombinedShannonEntropy_SSE2;
   VP8LVectorMismatch = VectorMismatch_SSE2;
   VP8LBundleColorMap = BundleColorMap_SSE2;

src/dsp/lossless_enc_sse41.c

@@ -18,6 +18,9 @@
 #include <smmintrin.h>
 #include "src/dsp/lossless.h"
 
+// For sign-extended multiplying constants, pre-shifted by 5:
+#define CST_5b(X)  (((int16_t)((uint16_t)(X) << 8)) >> 5)
+
 //------------------------------------------------------------------------------
 // Subtract-Green Transform
 
@@ -38,6 +41,95 @@ static void SubtractGreenFromBlueAndRed_SSE41(uint32_t* argb_data,
   }
 }
 
+//------------------------------------------------------------------------------
+// Color Transform
+
+#define SPAN 8
+static void CollectColorBlueTransforms_SSE41(const uint32_t* argb, int stride,
+                                             int tile_width, int tile_height,
+                                             int green_to_blue, int red_to_blue,
+                                             int histo[]) {
+  const __m128i mults_r = _mm_set1_epi16(CST_5b(red_to_blue));
+  const __m128i mults_g = _mm_set1_epi16(CST_5b(green_to_blue));
+  const __m128i mask_g = _mm_set1_epi16(0xff00);   // green mask
+  const __m128i mask_gb = _mm_set1_epi32(0xffff);  // green/blue mask
+  const __m128i mask_b = _mm_set1_epi16(0x00ff);   // blue mask
+  const __m128i shuffler_lo = _mm_setr_epi8(-1, 2, -1, 6, -1, 10, -1, 14, -1,
+                                            -1, -1, -1, -1, -1, -1, -1);
+  const __m128i shuffler_hi = _mm_setr_epi8(-1, -1, -1, -1, -1, -1, -1, -1, -1,
+                                            2, -1, 6, -1, 10, -1, 14);
+  int y;
+  for (y = 0; y < tile_height; ++y) {
+    const uint32_t* const src = argb + y * stride;
+    int i, x;
+    for (x = 0; x + SPAN <= tile_width; x += SPAN) {
+      uint16_t values[SPAN];
+      const __m128i in0 = _mm_loadu_si128((__m128i*)&src[x + 0]);
+      const __m128i in1 = _mm_loadu_si128((__m128i*)&src[x + SPAN / 2]);
+      const __m128i r0 = _mm_shuffle_epi8(in0, shuffler_lo);
+      const __m128i r1 = _mm_shuffle_epi8(in1, shuffler_hi);
+      const __m128i r = _mm_or_si128(r0, r1);         // r 0
+      const __m128i gb0 = _mm_and_si128(in0, mask_gb);
+      const __m128i gb1 = _mm_and_si128(in1, mask_gb);
+      const __m128i gb = _mm_packus_epi32(gb0, gb1);  // g b
+      const __m128i g = _mm_and_si128(gb, mask_g);    // g 0
+      const __m128i A = _mm_mulhi_epi16(r, mults_r);  // x dbr
+      const __m128i B = _mm_mulhi_epi16(g, mults_g);  // x dbg
+      const __m128i C = _mm_sub_epi8(gb, B);          // x b'
+      const __m128i D = _mm_sub_epi8(C, A);           // x b''
+      const __m128i E = _mm_and_si128(D, mask_b);     // 0 b''
+      _mm_storeu_si128((__m128i*)values, E);
+      for (i = 0; i < SPAN; ++i) ++histo[values[i]];
+    }
+  }
+  {
+    const int left_over = tile_width & (SPAN - 1);
+    if (left_over > 0) {
+      VP8LCollectColorBlueTransforms_C(argb + tile_width - left_over, stride,
+                                       left_over, tile_height,
+                                       green_to_blue, red_to_blue, histo);
+    }
+  }
+}
+
+static void CollectColorRedTransforms_SSE41(const uint32_t* argb, int stride,
+                                            int tile_width, int tile_height,
+                                            int green_to_red, int histo[]) {
+  const __m128i mults_g = _mm_set1_epi16(CST_5b(green_to_red));
+  const __m128i mask_g = _mm_set1_epi32(0x00ff00);  // green mask
+  const __m128i mask = _mm_set1_epi16(0xff);
+
+  int y;
+  for (y = 0; y < tile_height; ++y) {
+    const uint32_t* const src = argb + y * stride;
+    int i, x;
+    for (x = 0; x + SPAN <= tile_width; x += SPAN) {
+      uint16_t values[SPAN];
+      const __m128i in0 = _mm_loadu_si128((__m128i*)&src[x + 0]);
+      const __m128i in1 = _mm_loadu_si128((__m128i*)&src[x + SPAN / 2]);
+      const __m128i g0 = _mm_and_si128(in0, mask_g);  // 0 0  | g 0
+      const __m128i g1 = _mm_and_si128(in1, mask_g);
+      const __m128i g = _mm_packus_epi32(g0, g1);     // g 0
+      const __m128i A0 = _mm_srli_epi32(in0, 16);     // 0 0  | x r
+      const __m128i A1 = _mm_srli_epi32(in1, 16);
+      const __m128i A = _mm_packus_epi32(A0, A1);     // x r
+      const __m128i B = _mm_mulhi_epi16(g, mults_g);  // x dr
+      const __m128i C = _mm_sub_epi8(A, B);           // x r'
+      const __m128i D = _mm_and_si128(C, mask);       // 0 r'
+      _mm_storeu_si128((__m128i*)values, D);
+      for (i = 0; i < SPAN; ++i) ++histo[values[i]];
+    }
+  }
+  {
+    const int left_over = tile_width & (SPAN - 1);
+    if (left_over > 0) {
+      VP8LCollectColorRedTransforms_C(argb + tile_width - left_over, stride,
+                                      left_over, tile_height, green_to_red,
+                                      histo);
+    }
+  }
+}
+
 //------------------------------------------------------------------------------
 // Entry point
 
@@ -45,6 +137,8 @@ extern void VP8LEncDspInitSSE41(void);
 
 WEBP_TSAN_IGNORE_FUNCTION void VP8LEncDspInitSSE41(void) {
   VP8LSubtractGreenFromBlueAndRed = SubtractGreenFromBlueAndRed_SSE41;
+  VP8LCollectColorBlueTransforms = CollectColorBlueTransforms_SSE41;
+  VP8LCollectColorRedTransforms = CollectColorRedTransforms_SSE41;
 }
 
 #else  // !WEBP_USE_SSE41

src/dsp/lossless_sse2.c

@@ -453,14 +453,11 @@ static void TransformColorInverse_SSE2(const VP8LMultipliers* const m,
                                        int num_pixels, uint32_t* dst) {
 // sign-extended multiplying constants, pre-shifted by 5.
 #define CST(X)  (((int16_t)(m->X << 8)) >> 5)   // sign-extend
-  const __m128i mults_rb = _mm_set_epi16(
-      CST(green_to_red_), CST(green_to_blue_),
-      CST(green_to_red_), CST(green_to_blue_),
-      CST(green_to_red_), CST(green_to_blue_),
-      CST(green_to_red_), CST(green_to_blue_));
-  const __m128i mults_b2 = _mm_set_epi16(
-      CST(red_to_blue_), 0, CST(red_to_blue_), 0,
-      CST(red_to_blue_), 0, CST(red_to_blue_), 0);
+#define MK_CST_16(HI, LO) \
+  _mm_set1_epi32((int)(((uint32_t)(HI) << 16) | ((LO) & 0xffff)))
+  const __m128i mults_rb = MK_CST_16(CST(green_to_red_), CST(green_to_blue_));
+  const __m128i mults_b2 = MK_CST_16(CST(red_to_blue_), 0);
+#undef MK_CST_16
 #undef CST
   const __m128i mask_ag = _mm_set1_epi32(0xff00ff00);  // alpha-green masks
   int i;
@@ -503,11 +500,11 @@ static void ConvertBGRAToRGB_SSE2(const uint32_t* src, int num_pixels,
     __m128i in5 = _mm_loadu_si128(in + 5);
     __m128i in6 = _mm_loadu_si128(in + 6);
     __m128i in7 = _mm_loadu_si128(in + 7);
-    VP8L32bToPlanar(&in0, &in1, &in2, &in3);
-    VP8L32bToPlanar(&in4, &in5, &in6, &in7);
+    VP8L32bToPlanar_SSE2(&in0, &in1, &in2, &in3);
+    VP8L32bToPlanar_SSE2(&in4, &in5, &in6, &in7);
     // At this points, in1/in5 contains red only, in2/in6 green only ...
     // Pack the colors in 24b RGB.
-    VP8PlanarTo24b(&in1, &in5, &in2, &in6, &in3, &in7);
+    VP8PlanarTo24b_SSE2(&in1, &in5, &in2, &in6, &in3, &in7);
     _mm_storeu_si128(out + 0, in1);
     _mm_storeu_si128(out + 1, in5);
     _mm_storeu_si128(out + 2, in2);

src/dsp/msa_macro.h

@@ -1389,4 +1389,4 @@ static WEBP_INLINE uint32_t func_hadd_uh_u32(v8u16 in) {
 } while (0)
 #define AVER_UB2_UB(...) AVER_UB2(v16u8, __VA_ARGS__)
 
-#endif  /* WEBP_DSP_MSA_MACRO_H_ */
+#endif  // WEBP_DSP_MSA_MACRO_H_

src/dsp/rescaler.c

@@ -21,6 +21,7 @@
 
 #define ROUNDER (WEBP_RESCALER_ONE >> 1)
 #define MULT_FIX(x, y) (((uint64_t)(x) * (y) + ROUNDER) >> WEBP_RESCALER_RFIX)
+#define MULT_FIX_FLOOR(x, y) (((uint64_t)(x) * (y)) >> WEBP_RESCALER_RFIX)
 
 //------------------------------------------------------------------------------
 // Row import
@@ -138,7 +139,7 @@ void WebPRescalerExportRowShrink_C(WebPRescaler* const wrk) {
   if (yscale) {
     for (x_out = 0; x_out < x_out_max; ++x_out) {
       const uint32_t frac = (uint32_t)MULT_FIX(frow[x_out], yscale);
-      const int v = (int)MULT_FIX(irow[x_out] - frac, wrk->fxy_scale);
+      const int v = (int)MULT_FIX_FLOOR(irow[x_out] - frac, wrk->fxy_scale);
       assert(v >= 0 && v <= 255);
       dst[x_out] = v;
       irow[x_out] = frac;   // new fractional start
@@ -153,6 +154,7 @@ void WebPRescalerExportRowShrink_C(WebPRescaler* const wrk) {
   }
 }
 
+#undef MULT_FIX_FLOOR
 #undef MULT_FIX
 #undef ROUNDER
 
@@ -204,11 +206,7 @@ extern void WebPRescalerDspInitMIPSdspR2(void);
 extern void WebPRescalerDspInitMSA(void);
 extern void WebPRescalerDspInitNEON(void);
 
-static volatile VP8CPUInfo rescaler_last_cpuinfo_used =
-    (VP8CPUInfo)&rescaler_last_cpuinfo_used;
-
-WEBP_TSAN_IGNORE_FUNCTION void WebPRescalerDspInit(void) {
-  if (rescaler_last_cpuinfo_used == VP8GetCPUInfo) return;
+WEBP_DSP_INIT_FUNC(WebPRescalerDspInit) {
 #if !defined(WEBP_REDUCE_SIZE)
 #if !WEBP_NEON_OMIT_C_CODE
   WebPRescalerExportRowExpand = WebPRescalerExportRowExpand_C;
@@ -253,5 +251,4 @@ WEBP_TSAN_IGNORE_FUNCTION void WebPRescalerDspInit(void) {
   assert(WebPRescalerImportRowExpand != NULL);
   assert(WebPRescalerImportRowShrink != NULL);
 #endif   // WEBP_REDUCE_SIZE
-  rescaler_last_cpuinfo_used = VP8GetCPUInfo;
 }

src/dsp/rescaler_mips32.c

@@ -209,6 +209,7 @@ static void ExportRowExpand_MIPS32(WebPRescaler* const wrk) {
   }
 }
 
+#if 0  // disabled for now. TODO(skal): make match the C-code
 static void ExportRowShrink_MIPS32(WebPRescaler* const wrk) {
   const int x_out_max = wrk->dst_width * wrk->num_channels;
   uint8_t* dst = wrk->dst;
@@ -273,6 +274,7 @@ static void ExportRowShrink_MIPS32(WebPRescaler* const wrk) {
     );
   }
 }
+#endif  // 0
 
 //------------------------------------------------------------------------------
 // Entry point
@@ -283,7 +285,7 @@ WEBP_TSAN_IGNORE_FUNCTION void WebPRescalerDspInitMIPS32(void) {
   WebPRescalerImportRowExpand = ImportRowExpand_MIPS32;
   WebPRescalerImportRowShrink = ImportRowShrink_MIPS32;
   WebPRescalerExportRowExpand = ExportRowExpand_MIPS32;
-  WebPRescalerExportRowShrink = ExportRowShrink_MIPS32;
+//  WebPRescalerExportRowShrink = ExportRowShrink_MIPS32;
 }
 
 #else  // !WEBP_USE_MIPS32

src/dsp/rescaler_mips_dsp_r2.c

@@ -20,10 +20,12 @@
 
 #define ROUNDER (WEBP_RESCALER_ONE >> 1)
 #define MULT_FIX(x, y) (((uint64_t)(x) * (y) + ROUNDER) >> WEBP_RESCALER_RFIX)
+#define MULT_FIX_FLOOR(x, y) (((uint64_t)(x) * (y)) >> WEBP_RESCALER_RFIX)
 
 //------------------------------------------------------------------------------
 // Row export
 
+#if 0  // disabled for now. TODO(skal): make match the C-code
 static void ExportRowShrink_MIPSdspR2(WebPRescaler* const wrk) {
   int i;
   const int x_out_max = wrk->dst_width * wrk->num_channels;
@@ -106,7 +108,7 @@ static void ExportRowShrink_MIPSdspR2(WebPRescaler* const wrk) {
     }
     for (i = 0; i < (x_out_max & 0x3); ++i) {
       const uint32_t frac = (uint32_t)MULT_FIX(*frow++, yscale);
-      const int v = (int)MULT_FIX(*irow - frac, wrk->fxy_scale);
+      const int v = (int)MULT_FIX_FLOOR(*irow - frac, wrk->fxy_scale);
       assert(v >= 0 && v <= 255);
       *dst++ = v;
       *irow++ = frac;   // new fractional start
@@ -154,13 +156,14 @@ static void ExportRowShrink_MIPSdspR2(WebPRescaler* const wrk) {
       );
     }
     for (i = 0; i < (x_out_max & 0x3); ++i) {
-      const int v = (int)MULT_FIX(*irow, wrk->fxy_scale);
+      const int v = (int)MULT_FIX_FLOOR(*irow, wrk->fxy_scale);
       assert(v >= 0 && v <= 255);
       *dst++ = v;
       *irow++ = 0;
     }
   }
 }
+#endif  // 0
 
 static void ExportRowExpand_MIPSdspR2(WebPRescaler* const wrk) {
   int i;
@@ -294,6 +297,7 @@ static void ExportRowExpand_MIPSdspR2(WebPRescaler* const wrk) {
   }
 }
 
+#undef MULT_FIX_FLOOR
 #undef MULT_FIX
 #undef ROUNDER
 
@@ -304,7 +308,7 @@ extern void WebPRescalerDspInitMIPSdspR2(void);
 
 WEBP_TSAN_IGNORE_FUNCTION void WebPRescalerDspInitMIPSdspR2(void) {
   WebPRescalerExportRowExpand = ExportRowExpand_MIPSdspR2;
-  WebPRescalerExportRowShrink = ExportRowShrink_MIPSdspR2;
+//  WebPRescalerExportRowShrink = ExportRowShrink_MIPSdspR2;
 }
 
 #else  // !WEBP_USE_MIPS_DSP_R2

src/dsp/rescaler_msa.c

@@ -22,6 +22,7 @@
 
 #define ROUNDER (WEBP_RESCALER_ONE >> 1)
 #define MULT_FIX(x, y) (((uint64_t)(x) * (y) + ROUNDER) >> WEBP_RESCALER_RFIX)
+#define MULT_FIX_FLOOR(x, y) (((uint64_t)(x) * (y)) >> WEBP_RESCALER_RFIX)
 
 #define CALC_MULT_FIX_16(in0, in1, in2, in3, scale, shift, dst) do {  \
   v4u32 tmp0, tmp1, tmp2, tmp3;                                       \
@@ -262,6 +263,7 @@ static void RescalerExportRowExpand_MIPSdspR2(WebPRescaler* const wrk) {
   }
 }
 
+#if 0  // disabled for now. TODO(skal): make match the C-code
 static WEBP_INLINE void ExportRowShrink_0(const uint32_t* frow, uint32_t* irow,
                                           uint8_t* dst, int length,
                                           const uint32_t yscale,
@@ -341,7 +343,7 @@ static WEBP_INLINE void ExportRowShrink_0(const uint32_t* frow, uint32_t* irow,
     }
     for (x_out = 0; x_out < length; ++x_out) {
       const uint32_t frac = (uint32_t)MULT_FIX(frow[x_out], yscale);
-      const int v = (int)MULT_FIX(irow[x_out] - frac, wrk->fxy_scale);
+      const int v = (int)MULT_FIX_FLOOR(irow[x_out] - frac, wrk->fxy_scale);
       assert(v >= 0 && v <= 255);
       dst[x_out] = v;
       irow[x_out] = frac;
@@ -426,6 +428,7 @@ static void RescalerExportRowShrink_MIPSdspR2(WebPRescaler* const wrk) {
     ExportRowShrink_1(irow, dst, x_out_max, wrk);
   }
 }
+#endif  // 0
 
 //------------------------------------------------------------------------------
 // Entry point
@@ -434,7 +437,7 @@ extern void WebPRescalerDspInitMSA(void);
 
 WEBP_TSAN_IGNORE_FUNCTION void WebPRescalerDspInitMSA(void) {
   WebPRescalerExportRowExpand = RescalerExportRowExpand_MIPSdspR2;
-  WebPRescalerExportRowShrink = RescalerExportRowShrink_MIPSdspR2;
+//  WebPRescalerExportRowShrink = RescalerExportRowShrink_MIPSdspR2;
 }
 
 #else     // !WEBP_USE_MSA

src/dsp/rescaler_neon.c

@@ -22,6 +22,7 @@
 
 #define ROUNDER (WEBP_RESCALER_ONE >> 1)
 #define MULT_FIX_C(x, y) (((uint64_t)(x) * (y) + ROUNDER) >> WEBP_RESCALER_RFIX)
+#define MULT_FIX_FLOOR_C(x, y) (((uint64_t)(x) * (y)) >> WEBP_RESCALER_RFIX)
 
 #define LOAD_32x4(SRC, DST) const uint32x4_t DST = vld1q_u32((SRC))
 #define LOAD_32x8(SRC, DST0, DST1)                                    \
@@ -35,8 +36,11 @@
 
 #if (WEBP_RESCALER_RFIX == 32)
 #define MAKE_HALF_CST(C) vdupq_n_s32((int32_t)((C) >> 1))
-#define MULT_FIX(A, B) /* note: B is actualy scale>>1. See MAKE_HALF_CST */ \
+// note: B is actualy scale>>1. See MAKE_HALF_CST
+#define MULT_FIX(A, B) \
     vreinterpretq_u32_s32(vqrdmulhq_s32(vreinterpretq_s32_u32((A)), (B)))
+#define MULT_FIX_FLOOR(A, B) \
+    vreinterpretq_u32_s32(vqdmulhq_s32(vreinterpretq_s32_u32((A)), (B)))
 #else
 #error "MULT_FIX/WEBP_RESCALER_RFIX need some more work"
 #endif
@@ -135,8 +139,8 @@ static void RescalerExportRowShrink_NEON(WebPRescaler* const wrk) {
       const uint32x4_t A1 = MULT_FIX(in1, yscale_half);
       const uint32x4_t B0 = vqsubq_u32(in2, A0);
       const uint32x4_t B1 = vqsubq_u32(in3, A1);
-      const uint32x4_t C0 = MULT_FIX(B0, fxy_scale_half);
-      const uint32x4_t C1 = MULT_FIX(B1, fxy_scale_half);
+      const uint32x4_t C0 = MULT_FIX_FLOOR(B0, fxy_scale_half);
+      const uint32x4_t C1 = MULT_FIX_FLOOR(B1, fxy_scale_half);
       const uint16x4_t D0 = vmovn_u32(C0);
       const uint16x4_t D1 = vmovn_u32(C1);
       const uint8x8_t E = vmovn_u16(vcombine_u16(D0, D1));
@@ -145,7 +149,7 @@ static void RescalerExportRowShrink_NEON(WebPRescaler* const wrk) {
     }
     for (; x_out < x_out_max; ++x_out) {
       const uint32_t frac = (uint32_t)MULT_FIX_C(frow[x_out], yscale);
-      const int v = (int)MULT_FIX_C(irow[x_out] - frac, wrk->fxy_scale);
+      const int v = (int)MULT_FIX_FLOOR_C(irow[x_out] - frac, fxy_scale);
       assert(v >= 0 && v <= 255);
       dst[x_out] = v;
       irow[x_out] = frac;   // new fractional start
@@ -170,6 +174,12 @@ static void RescalerExportRowShrink_NEON(WebPRescaler* const wrk) {
   }
 }
 
+#undef MULT_FIX_FLOOR_C
+#undef MULT_FIX_C
+#undef MULT_FIX_FLOOR
+#undef MULT_FIX
+#undef ROUNDER
+
 //------------------------------------------------------------------------------
 
 extern void WebPRescalerDspInitNEON(void);

src/dsp/rescaler_sse2.c

@@ -25,6 +25,7 @@
 
 #define ROUNDER (WEBP_RESCALER_ONE >> 1)
 #define MULT_FIX(x, y) (((uint64_t)(x) * (y) + ROUNDER) >> WEBP_RESCALER_RFIX)
+#define MULT_FIX_FLOOR(x, y) (((uint64_t)(x) * (y)) >> WEBP_RESCALER_RFIX)
 
 // input: 8 bytes ABCDEFGH -> output: A0E0B0F0C0G0D0H0
 static void LoadTwoPixels_SSE2(const uint8_t* const src, __m128i* out) {
@@ -36,7 +37,7 @@ static void LoadTwoPixels_SSE2(const uint8_t* const src, __m128i* out) {
 }
 
 // input: 8 bytes ABCDEFGH -> output: A0B0C0D0E0F0G0H0
-static void LoadHeightPixels_SSE2(const uint8_t* const src, __m128i* out) {
+static void LoadEightPixels_SSE2(const uint8_t* const src, __m128i* out) {
   const __m128i zero = _mm_setzero_si128();
   const __m128i A = _mm_loadl_epi64((const __m128i*)(src));  // ABCDEFGH
   *out = _mm_unpacklo_epi8(A, zero);
@@ -50,13 +51,15 @@ static void RescalerImportRowExpand_SSE2(WebPRescaler* const wrk,
   int accum = x_add;
   __m128i cur_pixels;
 
+  // SSE2 implementation only works with 16b signed arithmetic at max.
+  if (wrk->src_width < 8 || accum >= (1 << 15)) {
+    WebPRescalerImportRowExpand_C(wrk, src);
+    return;
+  }
+
   assert(!WebPRescalerInputDone(wrk));
   assert(wrk->x_expand);
   if (wrk->num_channels == 4) {
-    if (wrk->src_width < 2) {
-      WebPRescalerImportRowExpand_C(wrk, src);
-      return;
-    }
     LoadTwoPixels_SSE2(src, &cur_pixels);
     src += 4;
     while (1) {
@@ -75,11 +78,7 @@ static void RescalerImportRowExpand_SSE2(WebPRescaler* const wrk,
   } else {
     int left;
     const uint8_t* const src_limit = src + wrk->src_width - 8;
-    if (wrk->src_width < 8) {
-      WebPRescalerImportRowExpand_C(wrk, src);
-      return;
-    }
-    LoadHeightPixels_SSE2(src, &cur_pixels);
+    LoadEightPixels_SSE2(src, &cur_pixels);
     src += 7;
     left = 7;
     while (1) {
@@ -94,7 +93,7 @@ static void RescalerImportRowExpand_SSE2(WebPRescaler* const wrk,
         if (--left) {
           cur_pixels = _mm_srli_si128(cur_pixels, 2);
         } else if (src <= src_limit) {
-          LoadHeightPixels_SSE2(src, &cur_pixels);
+          LoadEightPixels_SSE2(src, &cur_pixels);
           src += 7;
           left = 7;
         } else {   // tail
@@ -226,6 +225,35 @@ static WEBP_INLINE void ProcessRow_SSE2(const __m128i* const A0,
   _mm_storel_epi64((__m128i*)dst, G);
 }
 
+static WEBP_INLINE void ProcessRow_Floor_SSE2(const __m128i* const A0,
+                                              const __m128i* const A1,
+                                              const __m128i* const A2,
+                                              const __m128i* const A3,
+                                              const __m128i* const mult,
+                                              uint8_t* const dst) {
+  const __m128i mask = _mm_set_epi32(0xffffffffu, 0, 0xffffffffu, 0);
+  const __m128i B0 = _mm_mul_epu32(*A0, *mult);
+  const __m128i B1 = _mm_mul_epu32(*A1, *mult);
+  const __m128i B2 = _mm_mul_epu32(*A2, *mult);
+  const __m128i B3 = _mm_mul_epu32(*A3, *mult);
+  const __m128i D0 = _mm_srli_epi64(B0, WEBP_RESCALER_RFIX);
+  const __m128i D1 = _mm_srli_epi64(B1, WEBP_RESCALER_RFIX);
+#if (WEBP_RESCALER_RFIX < 32)
+  const __m128i D2 =
+      _mm_and_si128(_mm_slli_epi64(B2, 32 - WEBP_RESCALER_RFIX), mask);
+  const __m128i D3 =
+      _mm_and_si128(_mm_slli_epi64(B3, 32 - WEBP_RESCALER_RFIX), mask);
+#else
+  const __m128i D2 = _mm_and_si128(B2, mask);
+  const __m128i D3 = _mm_and_si128(B3, mask);
+#endif
+  const __m128i E0 = _mm_or_si128(D0, D2);
+  const __m128i E1 = _mm_or_si128(D1, D3);
+  const __m128i F = _mm_packs_epi32(E0, E1);
+  const __m128i G = _mm_packus_epi16(F, F);
+  _mm_storel_epi64((__m128i*)dst, G);
+}
+
 static void RescalerExportRowExpand_SSE2(WebPRescaler* const wrk) {
   int x_out;
   uint8_t* const dst = wrk->dst;
@@ -324,12 +352,12 @@ static void RescalerExportRowShrink_SSE2(WebPRescaler* const wrk) {
         const __m128i G1 = _mm_or_si128(D1, F3);
         _mm_storeu_si128((__m128i*)(irow + x_out + 0), G0);
         _mm_storeu_si128((__m128i*)(irow + x_out + 4), G1);
-        ProcessRow_SSE2(&E0, &E1, &E2, &E3, &mult_xy, dst + x_out);
+        ProcessRow_Floor_SSE2(&E0, &E1, &E2, &E3, &mult_xy, dst + x_out);
       }
     }
     for (; x_out < x_out_max; ++x_out) {
       const uint32_t frac = (int)MULT_FIX(frow[x_out], yscale);
-      const int v = (int)MULT_FIX(irow[x_out] - frac, wrk->fxy_scale);
+      const int v = (int)MULT_FIX_FLOOR(irow[x_out] - frac, wrk->fxy_scale);
       assert(v >= 0 && v <= 255);
       dst[x_out] = v;
       irow[x_out] = frac;   // new fractional start
@@ -354,6 +382,7 @@ static void RescalerExportRowShrink_SSE2(WebPRescaler* const wrk) {
   }
 }
 
+#undef MULT_FIX_FLOOR
 #undef MULT_FIX
 #undef ROUNDER
 

src/dsp/ssim.c

@@ -139,12 +139,7 @@ VP8AccumulateSSEFunc VP8AccumulateSSE;
 
 extern void VP8SSIMDspInitSSE2(void);
 
-static volatile VP8CPUInfo ssim_last_cpuinfo_used =
-    (VP8CPUInfo)&ssim_last_cpuinfo_used;
-
-WEBP_TSAN_IGNORE_FUNCTION void VP8SSIMDspInit(void) {
-  if (ssim_last_cpuinfo_used == VP8GetCPUInfo) return;
-
+WEBP_DSP_INIT_FUNC(VP8SSIMDspInit) {
 #if !defined(WEBP_REDUCE_SIZE)
   VP8SSIMGetClipped = SSIMGetClipped_C;
   VP8SSIMGet = SSIMGet_C;
@@ -161,6 +156,4 @@ WEBP_TSAN_IGNORE_FUNCTION void VP8SSIMDspInit(void) {
     }
 #endif
   }
-
-  ssim_last_cpuinfo_used = VP8GetCPUInfo;
 }

src/dsp/upsampling.c

@@ -217,13 +217,9 @@ WebPYUV444Converter WebPYUV444Converters[MODE_LAST];
 
 extern void WebPInitYUV444ConvertersMIPSdspR2(void);
 extern void WebPInitYUV444ConvertersSSE2(void);
+extern void WebPInitYUV444ConvertersSSE41(void);
 
-static volatile VP8CPUInfo upsampling_last_cpuinfo_used1 =
-    (VP8CPUInfo)&upsampling_last_cpuinfo_used1;
-
-WEBP_TSAN_IGNORE_FUNCTION void WebPInitYUV444Converters(void) {
-  if (upsampling_last_cpuinfo_used1 == VP8GetCPUInfo) return;
-
+WEBP_DSP_INIT_FUNC(WebPInitYUV444Converters) {
   WebPYUV444Converters[MODE_RGBA]      = WebPYuv444ToRgba_C;
   WebPYUV444Converters[MODE_BGRA]      = WebPYuv444ToBgra_C;
   WebPYUV444Converters[MODE_RGB]       = WebPYuv444ToRgb_C;
@@ -242,29 +238,29 @@ WEBP_TSAN_IGNORE_FUNCTION void WebPInitYUV444Converters(void) {
       WebPInitYUV444ConvertersSSE2();
     }
 #endif
+#if defined(WEBP_USE_SSE41)
+    if (VP8GetCPUInfo(kSSE4_1)) {
+      WebPInitYUV444ConvertersSSE41();
+    }
+#endif
 #if defined(WEBP_USE_MIPS_DSP_R2)
     if (VP8GetCPUInfo(kMIPSdspR2)) {
       WebPInitYUV444ConvertersMIPSdspR2();
     }
 #endif
   }
-  upsampling_last_cpuinfo_used1 = VP8GetCPUInfo;
 }
 
 //------------------------------------------------------------------------------
 // Main calls
 
 extern void WebPInitUpsamplersSSE2(void);
+extern void WebPInitUpsamplersSSE41(void);
 extern void WebPInitUpsamplersNEON(void);
 extern void WebPInitUpsamplersMIPSdspR2(void);
 extern void WebPInitUpsamplersMSA(void);
 
-static volatile VP8CPUInfo upsampling_last_cpuinfo_used2 =
-    (VP8CPUInfo)&upsampling_last_cpuinfo_used2;
-
-WEBP_TSAN_IGNORE_FUNCTION void WebPInitUpsamplers(void) {
-  if (upsampling_last_cpuinfo_used2 == VP8GetCPUInfo) return;
-
+WEBP_DSP_INIT_FUNC(WebPInitUpsamplers) {
 #ifdef FANCY_UPSAMPLING
 #if !WEBP_NEON_OMIT_C_CODE
   WebPUpsamplers[MODE_RGBA]      = UpsampleRgbaLinePair_C;
@@ -287,6 +283,11 @@ WEBP_TSAN_IGNORE_FUNCTION void WebPInitUpsamplers(void) {
       WebPInitUpsamplersSSE2();
     }
 #endif
+#if defined(WEBP_USE_SSE41)
+    if (VP8GetCPUInfo(kSSE4_1)) {
+      WebPInitUpsamplersSSE41();
+    }
+#endif
 #if defined(WEBP_USE_MIPS_DSP_R2)
     if (VP8GetCPUInfo(kMIPSdspR2)) {
       WebPInitUpsamplersMIPSdspR2();
@@ -310,6 +311,7 @@ WEBP_TSAN_IGNORE_FUNCTION void WebPInitUpsamplers(void) {
   assert(WebPUpsamplers[MODE_BGRA] != NULL);
   assert(WebPUpsamplers[MODE_rgbA] != NULL);
   assert(WebPUpsamplers[MODE_bgrA] != NULL);
+#if !defined(WEBP_REDUCE_CSP) || !WEBP_NEON_OMIT_C_CODE
   assert(WebPUpsamplers[MODE_RGB] != NULL);
   assert(WebPUpsamplers[MODE_BGR] != NULL);
   assert(WebPUpsamplers[MODE_ARGB] != NULL);
@@ -317,9 +319,9 @@ WEBP_TSAN_IGNORE_FUNCTION void WebPInitUpsamplers(void) {
   assert(WebPUpsamplers[MODE_RGB_565] != NULL);
   assert(WebPUpsamplers[MODE_Argb] != NULL);
   assert(WebPUpsamplers[MODE_rgbA_4444] != NULL);
+#endif
 
 #endif  // FANCY_UPSAMPLING
-  upsampling_last_cpuinfo_used2 = VP8GetCPUInfo;
 }
 
 //------------------------------------------------------------------------------

src/dsp/upsampling_msa.c

@@ -264,6 +264,7 @@ static void YuvToBgr(int y, int u, int v, uint8_t* const bgr) {
   bgr[2] = Clip8(r1 >> 6);
 }
 
+#if !defined(WEBP_REDUCE_CSP)
 static void YuvToRgb565(int y, int u, int v, uint8_t* const rgb) {
   const int y1 = MultHi(y, 19077);
   const int r1 = y1 + MultHi(v, 26149) - 14234;
@@ -306,6 +307,7 @@ static void YuvToArgb(uint8_t y, uint8_t u, uint8_t v, uint8_t* const argb) {
   argb[0] = 0xff;
   YuvToRgb(y, u, v, argb + 1);
 }
+#endif  // WEBP_REDUCE_CSP
 
 static void YuvToBgra(uint8_t y, uint8_t u, uint8_t v, uint8_t* const bgra) {
   YuvToBgr(y, u, v, bgra);
@@ -317,6 +319,7 @@ static void YuvToRgba(uint8_t y, uint8_t u, uint8_t v, uint8_t* const rgba) {
   rgba[3] = 0xff;
 }
 
+#if !defined(WEBP_REDUCE_CSP)
 static void YuvToRgbLine(const uint8_t* y, const uint8_t* u,
                          const uint8_t* v, uint8_t* dst, int length) {
   v16u8 R, G, B;
@@ -370,6 +373,7 @@ static void YuvToBgrLine(const uint8_t* y, const uint8_t* u,
     memcpy(dst, temp, length * 3 * sizeof(*dst));
   }
 }
+#endif  // WEBP_REDUCE_CSP
 
 static void YuvToRgbaLine(const uint8_t* y, const uint8_t* u,
                           const uint8_t* v, uint8_t* dst, int length) {
@@ -427,6 +431,7 @@ static void YuvToBgraLine(const uint8_t* y, const uint8_t* u,
   }
 }
 
+#if !defined(WEBP_REDUCE_CSP)
 static void YuvToArgbLine(const uint8_t* y, const uint8_t* u,
                           const uint8_t* v, uint8_t* dst, int length) {
   v16u8 R, G, B;
@@ -526,6 +531,7 @@ static void YuvToRgb565Line(const uint8_t* y, const uint8_t* u,
     memcpy(dst, temp, length * 2 * sizeof(*dst));
   }
 }
+#endif  // WEBP_REDUCE_CSP
 
 #define UPSAMPLE_32PIXELS(a, b, c, d) do {    \
   v16u8 s = __msa_aver_u_b(a, d);             \

src/dsp/upsampling_sse2.c

@@ -104,21 +104,6 @@ static void Upsample32Pixels_SSE2(const uint8_t r1[], const uint8_t r2[],
   Upsample32Pixels_SSE2(r1, r2, out);                                          \
 }
 
-#define CONVERT2RGB(FUNC, XSTEP, top_y, bottom_y,                              \
-                    top_dst, bottom_dst, cur_x, num_pixels) {                  \
-  int n;                                                                       \
-  for (n = 0; n < (num_pixels); ++n) {                                         \
-    FUNC((top_y)[(cur_x) + n], r_u[n], r_v[n],                                 \
-         (top_dst) + ((cur_x) + n) * (XSTEP));                                 \
-  }                                                                            \
-  if ((bottom_y) != NULL) {                                                    \
-    for (n = 0; n < (num_pixels); ++n) {                                       \
-      FUNC((bottom_y)[(cur_x) + n], r_u[64 + n], r_v[64 + n],                  \
-           (bottom_dst) + ((cur_x) + n) * (XSTEP));                            \
-    }                                                                          \
-  }                                                                            \
-}
-
 #define CONVERT2RGB_32(FUNC, XSTEP, top_y, bottom_y,                           \
                        top_dst, bottom_dst, cur_x) do {                        \
   FUNC##32_SSE2((top_y) + (cur_x), r_u, r_v, (top_dst) + (cur_x) * (XSTEP));   \
@@ -135,7 +120,7 @@ static void FUNC_NAME(const uint8_t* top_y, const uint8_t* bottom_y,           \
                       uint8_t* top_dst, uint8_t* bottom_dst, int len) {        \
   int uv_pos, pos;                                                             \
   /* 16byte-aligned array to cache reconstructed u and v */                    \
-  uint8_t uv_buf[4 * 32 + 15];                                                 \
+  uint8_t uv_buf[14 * 32 + 15] = { 0 };                                        \
   uint8_t* const r_u = (uint8_t*)((uintptr_t)(uv_buf + 15) & ~15);             \
   uint8_t* const r_v = r_u + 32;                                               \
                                                                                \
@@ -160,11 +145,22 @@ static void FUNC_NAME(const uint8_t* top_y, const uint8_t* bottom_y,           \
   }                                                                            \
   if (len > 1) {                                                               \
     const int left_over = ((len + 1) >> 1) - (pos >> 1);                       \
+    uint8_t* const tmp_top_dst = r_u + 4 * 32;                                 \
+    uint8_t* const tmp_bottom_dst = tmp_top_dst + 4 * 32;                      \
+    uint8_t* const tmp_top = tmp_bottom_dst + 4 * 32;                          \
+    uint8_t* const tmp_bottom = (bottom_y == NULL) ? NULL : tmp_top + 32;      \
     assert(left_over > 0);                                                     \
     UPSAMPLE_LAST_BLOCK(top_u + uv_pos, cur_u + uv_pos, left_over, r_u);       \
     UPSAMPLE_LAST_BLOCK(top_v + uv_pos, cur_v + uv_pos, left_over, r_v);       \
-    CONVERT2RGB(FUNC, XSTEP, top_y, bottom_y, top_dst, bottom_dst,             \
-                pos, len - pos);                                               \
+    memcpy(tmp_top, top_y + pos, len - pos);                                   \
+    if (bottom_y != NULL) memcpy(tmp_bottom, bottom_y + pos, len - pos);       \
+    CONVERT2RGB_32(FUNC, XSTEP, tmp_top, tmp_bottom, tmp_top_dst,              \
+         tmp_bottom_dst, 0);                                                   \
+    memcpy(top_dst + pos * (XSTEP), tmp_top_dst, (len - pos) * (XSTEP));       \
+    if (bottom_y != NULL) {                                                    \
+      memcpy(bottom_dst + pos * (XSTEP), tmp_bottom_dst,                       \
+             (len - pos) * (XSTEP));                                           \
+    }                                                                          \
   }                                                                            \
 }
 

src/dsp/upsampling_sse41.c

@@ -0,0 +1,239 @@
+// Copyright 2011 Google Inc. All Rights Reserved.
+//
+// Use of this source code is governed by a BSD-style license
+// that can be found in the COPYING file in the root of the source
+// tree. An additional intellectual property rights grant can be found
+// in the file PATENTS. All contributing project authors may
+// be found in the AUTHORS file in the root of the source tree.
+// -----------------------------------------------------------------------------
+//
+// SSE41 version of YUV to RGB upsampling functions.
+//
+// Author: somnath@google.com (Somnath Banerjee)
+
+#include "src/dsp/dsp.h"
+
+#if defined(WEBP_USE_SSE41)
+
+#include <assert.h>
+#include <smmintrin.h>
+#include <string.h>
+#include "src/dsp/yuv.h"
+
+#ifdef FANCY_UPSAMPLING
+
+#if !defined(WEBP_REDUCE_CSP)
+
+// We compute (9*a + 3*b + 3*c + d + 8) / 16 as follows
+// u = (9*a + 3*b + 3*c + d + 8) / 16
+//   = (a + (a + 3*b + 3*c + d) / 8 + 1) / 2
+//   = (a + m + 1) / 2
+// where m = (a + 3*b + 3*c + d) / 8
+//         = ((a + b + c + d) / 2 + b + c) / 4
+//
+// Let's say  k = (a + b + c + d) / 4.
+// We can compute k as
+// k = (s + t + 1) / 2 - ((a^d) | (b^c) | (s^t)) & 1
+// where s = (a + d + 1) / 2 and t = (b + c + 1) / 2
+//
+// Then m can be written as
+// m = (k + t + 1) / 2 - (((b^c) & (s^t)) | (k^t)) & 1
+
+// Computes out = (k + in + 1) / 2 - ((ij & (s^t)) | (k^in)) & 1
+#define GET_M(ij, in, out) do {                                                \
+  const __m128i tmp0 = _mm_avg_epu8(k, (in));     /* (k + in + 1) / 2 */       \
+  const __m128i tmp1 = _mm_and_si128((ij), st);   /* (ij) & (s^t) */           \
+  const __m128i tmp2 = _mm_xor_si128(k, (in));    /* (k^in) */                 \
+  const __m128i tmp3 = _mm_or_si128(tmp1, tmp2);  /* ((ij) & (s^t)) | (k^in) */\
+  const __m128i tmp4 = _mm_and_si128(tmp3, one);  /* & 1 -> lsb_correction */  \
+  (out) = _mm_sub_epi8(tmp0, tmp4);    /* (k + in + 1) / 2 - lsb_correction */ \
+} while (0)
+
+// pack and store two alternating pixel rows
+#define PACK_AND_STORE(a, b, da, db, out) do {                                 \
+  const __m128i t_a = _mm_avg_epu8(a, da);  /* (9a + 3b + 3c +  d + 8) / 16 */ \
+  const __m128i t_b = _mm_avg_epu8(b, db);  /* (3a + 9b +  c + 3d + 8) / 16 */ \
+  const __m128i t_1 = _mm_unpacklo_epi8(t_a, t_b);                             \
+  const __m128i t_2 = _mm_unpackhi_epi8(t_a, t_b);                             \
+  _mm_store_si128(((__m128i*)(out)) + 0, t_1);                                 \
+  _mm_store_si128(((__m128i*)(out)) + 1, t_2);                                 \
+} while (0)
+
+// Loads 17 pixels each from rows r1 and r2 and generates 32 pixels.
+#define UPSAMPLE_32PIXELS(r1, r2, out) {                                       \
+  const __m128i one = _mm_set1_epi8(1);                                        \
+  const __m128i a = _mm_loadu_si128((const __m128i*)&(r1)[0]);                 \
+  const __m128i b = _mm_loadu_si128((const __m128i*)&(r1)[1]);                 \
+  const __m128i c = _mm_loadu_si128((const __m128i*)&(r2)[0]);                 \
+  const __m128i d = _mm_loadu_si128((const __m128i*)&(r2)[1]);                 \
+                                                                               \
+  const __m128i s = _mm_avg_epu8(a, d);        /* s = (a + d + 1) / 2 */       \
+  const __m128i t = _mm_avg_epu8(b, c);        /* t = (b + c + 1) / 2 */       \
+  const __m128i st = _mm_xor_si128(s, t);      /* st = s^t */                  \
+                                                                               \
+  const __m128i ad = _mm_xor_si128(a, d);      /* ad = a^d */                  \
+  const __m128i bc = _mm_xor_si128(b, c);      /* bc = b^c */                  \
+                                                                               \
+  const __m128i t1 = _mm_or_si128(ad, bc);     /* (a^d) | (b^c) */             \
+  const __m128i t2 = _mm_or_si128(t1, st);     /* (a^d) | (b^c) | (s^t) */     \
+  const __m128i t3 = _mm_and_si128(t2, one);   /* (a^d) | (b^c) | (s^t) & 1 */ \
+  const __m128i t4 = _mm_avg_epu8(s, t);                                       \
+  const __m128i k = _mm_sub_epi8(t4, t3);      /* k = (a + b + c + d) / 4 */   \
+  __m128i diag1, diag2;                                                        \
+                                                                               \
+  GET_M(bc, t, diag1);                  /* diag1 = (a + 3b + 3c + d) / 8 */    \
+  GET_M(ad, s, diag2);                  /* diag2 = (3a + b + c + 3d) / 8 */    \
+                                                                               \
+  /* pack the alternate pixels */                                              \
+  PACK_AND_STORE(a, b, diag1, diag2, (out) +      0);  /* store top */         \
+  PACK_AND_STORE(c, d, diag2, diag1, (out) + 2 * 32);  /* store bottom */      \
+}
+
+// Turn the macro into a function for reducing code-size when non-critical
+static void Upsample32Pixels_SSE41(const uint8_t r1[], const uint8_t r2[],
+                                  uint8_t* const out) {
+  UPSAMPLE_32PIXELS(r1, r2, out);
+}
+
+#define UPSAMPLE_LAST_BLOCK(tb, bb, num_pixels, out) {                         \
+  uint8_t r1[17], r2[17];                                                      \
+  memcpy(r1, (tb), (num_pixels));                                              \
+  memcpy(r2, (bb), (num_pixels));                                              \
+  /* replicate last byte */                                                    \
+  memset(r1 + (num_pixels), r1[(num_pixels) - 1], 17 - (num_pixels));          \
+  memset(r2 + (num_pixels), r2[(num_pixels) - 1], 17 - (num_pixels));          \
+  /* using the shared function instead of the macro saves ~3k code size */     \
+  Upsample32Pixels_SSE41(r1, r2, out);                                         \
+}
+
+#define CONVERT2RGB_32(FUNC, XSTEP, top_y, bottom_y,                           \
+                       top_dst, bottom_dst, cur_x) do {                        \
+  FUNC##32_SSE41((top_y) + (cur_x), r_u, r_v, (top_dst) + (cur_x) * (XSTEP));  \
+  if ((bottom_y) != NULL) {                                                    \
+    FUNC##32_SSE41((bottom_y) + (cur_x), r_u + 64, r_v + 64,                   \
+                  (bottom_dst) + (cur_x) * (XSTEP));                           \
+  }                                                                            \
+} while (0)
+
+#define SSE4_UPSAMPLE_FUNC(FUNC_NAME, FUNC, XSTEP)                             \
+static void FUNC_NAME(const uint8_t* top_y, const uint8_t* bottom_y,           \
+                      const uint8_t* top_u, const uint8_t* top_v,              \
+                      const uint8_t* cur_u, const uint8_t* cur_v,              \
+                      uint8_t* top_dst, uint8_t* bottom_dst, int len) {        \
+  int uv_pos, pos;                                                             \
+  /* 16byte-aligned array to cache reconstructed u and v */                    \
+  uint8_t uv_buf[14 * 32 + 15] = { 0 };                                        \
+  uint8_t* const r_u = (uint8_t*)((uintptr_t)(uv_buf + 15) & ~15);             \
+  uint8_t* const r_v = r_u + 32;                                               \
+                                                                               \
+  assert(top_y != NULL);                                                       \
+  {   /* Treat the first pixel in regular way */                               \
+    const int u_diag = ((top_u[0] + cur_u[0]) >> 1) + 1;                       \
+    const int v_diag = ((top_v[0] + cur_v[0]) >> 1) + 1;                       \
+    const int u0_t = (top_u[0] + u_diag) >> 1;                                 \
+    const int v0_t = (top_v[0] + v_diag) >> 1;                                 \
+    FUNC(top_y[0], u0_t, v0_t, top_dst);                                       \
+    if (bottom_y != NULL) {                                                    \
+      const int u0_b = (cur_u[0] + u_diag) >> 1;                               \
+      const int v0_b = (cur_v[0] + v_diag) >> 1;                               \
+      FUNC(bottom_y[0], u0_b, v0_b, bottom_dst);                               \
+    }                                                                          \
+  }                                                                            \
+  /* For UPSAMPLE_32PIXELS, 17 u/v values must be read-able for each block */  \
+  for (pos = 1, uv_pos = 0; pos + 32 + 1 <= len; pos += 32, uv_pos += 16) {    \
+    UPSAMPLE_32PIXELS(top_u + uv_pos, cur_u + uv_pos, r_u);                    \
+    UPSAMPLE_32PIXELS(top_v + uv_pos, cur_v + uv_pos, r_v);                    \
+    CONVERT2RGB_32(FUNC, XSTEP, top_y, bottom_y, top_dst, bottom_dst, pos);    \
+  }                                                                            \
+  if (len > 1) {                                                               \
+    const int left_over = ((len + 1) >> 1) - (pos >> 1);                       \
+    uint8_t* const tmp_top_dst = r_u + 4 * 32;                                 \
+    uint8_t* const tmp_bottom_dst = tmp_top_dst + 4 * 32;                      \
+    uint8_t* const tmp_top = tmp_bottom_dst + 4 * 32;                          \
+    uint8_t* const tmp_bottom = (bottom_y == NULL) ? NULL : tmp_top + 32;      \
+    assert(left_over > 0);                                                     \
+    UPSAMPLE_LAST_BLOCK(top_u + uv_pos, cur_u + uv_pos, left_over, r_u);       \
+    UPSAMPLE_LAST_BLOCK(top_v + uv_pos, cur_v + uv_pos, left_over, r_v);       \
+    memcpy(tmp_top, top_y + pos, len - pos);                                   \
+    if (bottom_y != NULL) memcpy(tmp_bottom, bottom_y + pos, len - pos);       \
+    CONVERT2RGB_32(FUNC, XSTEP, tmp_top, tmp_bottom, tmp_top_dst,              \
+         tmp_bottom_dst, 0);                                                   \
+    memcpy(top_dst + pos * (XSTEP), tmp_top_dst, (len - pos) * (XSTEP));       \
+    if (bottom_y != NULL) {                                                    \
+      memcpy(bottom_dst + pos * (XSTEP), tmp_bottom_dst,                       \
+             (len - pos) * (XSTEP));                                           \
+    }                                                                          \
+  }                                                                            \
+}
+
+// SSE4 variants of the fancy upsampler.
+SSE4_UPSAMPLE_FUNC(UpsampleRgbLinePair_SSE41,  VP8YuvToRgb,  3)
+SSE4_UPSAMPLE_FUNC(UpsampleBgrLinePair_SSE41,  VP8YuvToBgr,  3)
+
+#undef GET_M
+#undef PACK_AND_STORE
+#undef UPSAMPLE_32PIXELS
+#undef UPSAMPLE_LAST_BLOCK
+#undef CONVERT2RGB
+#undef CONVERT2RGB_32
+#undef SSE4_UPSAMPLE_FUNC
+
+#endif   // WEBP_REDUCE_CSP
+
+//------------------------------------------------------------------------------
+// Entry point
+
+extern WebPUpsampleLinePairFunc WebPUpsamplers[/* MODE_LAST */];
+
+extern void WebPInitUpsamplersSSE41(void);
+
+WEBP_TSAN_IGNORE_FUNCTION void WebPInitUpsamplersSSE41(void) {
+#if !defined(WEBP_REDUCE_CSP)
+  WebPUpsamplers[MODE_RGB]  = UpsampleRgbLinePair_SSE41;
+  WebPUpsamplers[MODE_BGR]  = UpsampleBgrLinePair_SSE41;
+#endif   // WEBP_REDUCE_CSP
+}
+
+#endif  // FANCY_UPSAMPLING
+
+//------------------------------------------------------------------------------
+
+extern WebPYUV444Converter WebPYUV444Converters[/* MODE_LAST */];
+extern void WebPInitYUV444ConvertersSSE41(void);
+
+#define YUV444_FUNC(FUNC_NAME, CALL, CALL_C, XSTEP)                            \
+extern void CALL_C(const uint8_t* y, const uint8_t* u, const uint8_t* v,       \
+                   uint8_t* dst, int len);                                     \
+static void FUNC_NAME(const uint8_t* y, const uint8_t* u, const uint8_t* v,    \
+                      uint8_t* dst, int len) {                                 \
+  int i;                                                                       \
+  const int max_len = len & ~31;                                               \
+  for (i = 0; i < max_len; i += 32) {                                          \
+    CALL(y + i, u + i, v + i, dst + i * (XSTEP));                              \
+  }                                                                            \
+  if (i < len) {  /* C-fallback */                                             \
+    CALL_C(y + i, u + i, v + i, dst + i * (XSTEP), len - i);                   \
+  }                                                                            \
+}
+
+#if !defined(WEBP_REDUCE_CSP)
+YUV444_FUNC(Yuv444ToRgb_SSE41, VP8YuvToRgb32_SSE41, WebPYuv444ToRgb_C, 3);
+YUV444_FUNC(Yuv444ToBgr_SSE41, VP8YuvToBgr32_SSE41, WebPYuv444ToBgr_C, 3);
+#endif  // WEBP_REDUCE_CSP
+
+WEBP_TSAN_IGNORE_FUNCTION void WebPInitYUV444ConvertersSSE41(void) {
+#if !defined(WEBP_REDUCE_CSP)
+  WebPYUV444Converters[MODE_RGB]       = Yuv444ToRgb_SSE41;
+  WebPYUV444Converters[MODE_BGR]       = Yuv444ToBgr_SSE41;
+#endif   // WEBP_REDUCE_CSP
+}
+
+#else
+
+WEBP_DSP_INIT_STUB(WebPInitYUV444ConvertersSSE41)
+
+#endif  // WEBP_USE_SSE41
+
+#if !(defined(FANCY_UPSAMPLING) && defined(WEBP_USE_SSE41))
+WEBP_DSP_INIT_STUB(WebPInitUpsamplersSSE41)
+#endif

src/dsp/yuv.c

@@ -71,15 +71,11 @@ void WebPSamplerProcessPlane(const uint8_t* y, int y_stride,
 WebPSamplerRowFunc WebPSamplers[MODE_LAST];
 
 extern void WebPInitSamplersSSE2(void);
+extern void WebPInitSamplersSSE41(void);
 extern void WebPInitSamplersMIPS32(void);
 extern void WebPInitSamplersMIPSdspR2(void);
 
-static volatile VP8CPUInfo yuv_last_cpuinfo_used =
-    (VP8CPUInfo)&yuv_last_cpuinfo_used;
-
-WEBP_TSAN_IGNORE_FUNCTION void WebPInitSamplers(void) {
-  if (yuv_last_cpuinfo_used == VP8GetCPUInfo) return;
-
+WEBP_DSP_INIT_FUNC(WebPInitSamplers) {
   WebPSamplers[MODE_RGB]       = YuvToRgbRow;
   WebPSamplers[MODE_RGBA]      = YuvToRgbaRow;
   WebPSamplers[MODE_BGR]       = YuvToBgrRow;
@@ -99,6 +95,11 @@ WEBP_TSAN_IGNORE_FUNCTION void WebPInitSamplers(void) {
       WebPInitSamplersSSE2();
     }
 #endif  // WEBP_USE_SSE2
+#if defined(WEBP_USE_SSE41)
+    if (VP8GetCPUInfo(kSSE4_1)) {
+      WebPInitSamplersSSE41();
+    }
+#endif  // WEBP_USE_SSE41
 #if defined(WEBP_USE_MIPS32)
     if (VP8GetCPUInfo(kMIPS32)) {
       WebPInitSamplersMIPS32();
@@ -110,7 +111,6 @@ WEBP_TSAN_IGNORE_FUNCTION void WebPInitSamplers(void) {
     }
 #endif  // WEBP_USE_MIPS_DSP_R2
   }
-  yuv_last_cpuinfo_used = VP8GetCPUInfo;
 }
 
 //-----------------------------------------------------------------------------
@@ -254,17 +254,13 @@ void (*WebPSharpYUVUpdateRGB)(const int16_t* ref, const int16_t* src,
 void (*WebPSharpYUVFilterRow)(const int16_t* A, const int16_t* B, int len,
                               const uint16_t* best_y, uint16_t* out);
 
-static volatile VP8CPUInfo rgba_to_yuv_last_cpuinfo_used =
-    (VP8CPUInfo)&rgba_to_yuv_last_cpuinfo_used;
-
 extern void WebPInitConvertARGBToYUVSSE2(void);
+extern void WebPInitConvertARGBToYUVSSE41(void);
 extern void WebPInitConvertARGBToYUVNEON(void);
 extern void WebPInitSharpYUVSSE2(void);
 extern void WebPInitSharpYUVNEON(void);
 
-WEBP_TSAN_IGNORE_FUNCTION void WebPInitConvertARGBToYUV(void) {
-  if (rgba_to_yuv_last_cpuinfo_used == VP8GetCPUInfo) return;
-
+WEBP_DSP_INIT_FUNC(WebPInitConvertARGBToYUV) {
   WebPConvertARGBToY = ConvertARGBToY_C;
   WebPConvertARGBToUV = WebPConvertARGBToUV_C;
 
@@ -286,6 +282,11 @@ WEBP_TSAN_IGNORE_FUNCTION void WebPInitConvertARGBToYUV(void) {
       WebPInitSharpYUVSSE2();
     }
 #endif  // WEBP_USE_SSE2
+#if defined(WEBP_USE_SSE41)
+    if (VP8GetCPUInfo(kSSE4_1)) {
+      WebPInitConvertARGBToYUVSSE41();
+    }
+#endif  // WEBP_USE_SSE41
   }
 
 #if defined(WEBP_USE_NEON)
@@ -304,6 +305,4 @@ WEBP_TSAN_IGNORE_FUNCTION void WebPInitConvertARGBToYUV(void) {
   assert(WebPSharpYUVUpdateY != NULL);
   assert(WebPSharpYUVUpdateRGB != NULL);
   assert(WebPSharpYUVFilterRow != NULL);
-
-  rgba_to_yuv_last_cpuinfo_used = VP8GetCPUInfo;
 }

src/dsp/yuv.h

@@ -166,6 +166,19 @@ void VP8YuvToRgb56532_SSE2(const uint8_t* y, const uint8_t* u, const uint8_t* v,
 
 #endif    // WEBP_USE_SSE2
 
+//-----------------------------------------------------------------------------
+// SSE41 extra functions (mostly for upsampling_sse41.c)
+
+#if defined(WEBP_USE_SSE41)
+
+// Process 32 pixels and store the result (16b, 24b or 32b per pixel) in *dst.
+void VP8YuvToRgb32_SSE41(const uint8_t* y, const uint8_t* u, const uint8_t* v,
+                         uint8_t* dst);
+void VP8YuvToBgr32_SSE41(const uint8_t* y, const uint8_t* u, const uint8_t* v,
+                         uint8_t* dst);
+
+#endif    // WEBP_USE_SSE41
+
 //------------------------------------------------------------------------------
 // RGB -> YUV conversion
 
@@ -194,4 +207,4 @@ static WEBP_INLINE int VP8RGBToV(int r, int g, int b, int rounding) {
 }    // extern "C"
 #endif
 
-#endif  /* WEBP_DSP_YUV_H_ */
+#endif  // WEBP_DSP_YUV_H_

src/dsp/yuv_sse2.c

@@ -180,7 +180,7 @@ static WEBP_INLINE void PlanarTo24b_SSE2(__m128i* const in0, __m128i* const in1,
   // Repeat the same permutations twice more:
   //   r0r4g0g4 | b0b4r1r5 | g1g5b1b5 | r2r6g2g6 | b2b6r3r7 | g3g7b3b7
   //   r0g0b0r1 | g1b1r2g2 | b2r3g3b3 | r4g4b4r5 | g5b5r6g6 | b6r7g7b7
-  VP8PlanarTo24b(in0, in1, in2, in3, in4, in5);
+  VP8PlanarTo24b_SSE2(in0, in1, in2, in3, in4, in5);
 
   _mm_storeu_si128((__m128i*)(rgb +  0), *in0);
   _mm_storeu_si128((__m128i*)(rgb + 16), *in1);
@@ -492,7 +492,7 @@ static WEBP_INLINE void RGB32PackedToPlanar_SSE2(const uint32_t* const argb,
   __m128i a1 = LOAD_16(argb + 4);
   __m128i a2 = LOAD_16(argb + 8);
   __m128i a3 = LOAD_16(argb + 12);
-  VP8L32bToPlanar(&a0, &a1, &a2, &a3);
+  VP8L32bToPlanar_SSE2(&a0, &a1, &a2, &a3);
   rgb[0] = _mm_unpacklo_epi8(a1, zero);
   rgb[1] = _mm_unpackhi_epi8(a1, zero);
   rgb[2] = _mm_unpacklo_epi8(a2, zero);